Gravel bed rivers VI: from process understanding to river restoration

Foreword The 6th International Gravel-Bed Rivers Workshop (GBR6) was held at St. Jakob/ Defereggental, near Lienz in Au...

Author: Helmut Habersack | Massimo Rinaldi | Hervé Piégay

42 downloads 897 Views 20MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Report copyright / DMCA form

DOWNLOAD PDF

Foreword The 6th International Gravel-Bed Rivers Workshop (GBR6) was held at St. Jakob/ Defereggental, near Lienz in Austria, between 5th and 9th September 2005. It was organised by a European scientiﬁc committee composed of Austrian, British, French, German, and Italian representatives with pre- and post-tours in the Austrian, French, German, and Italian Alps. This workshop was designed in continuity with previous GBR meetings, open to invited scientists and post-graduate students, with the goal of providing a forum for the review and discussion of research and developments in the preceding ﬁve years of all aspects of gravel-bed river sciences. Previous Gravel-Bed Rivers Workshops took place in:

England, Gregynog, Newtown, Wales, June 1980; USA, Pingree Park, Colorado, August 1985; Italy, Poggio a Caiano, September 1990; USA, Gold Bar, Washington, August 1995; and New Zealand, Christchurch, August–September 2000.

The subject of GBR6, ‘‘From process understanding to river restoration’’, is concerned with recent progress in the understanding of gravel-bed river morphology and sediment transport, as well as new developments in restoration. There was a particular emphasis on scaling aspects of gravel-bed river processes and patterns. GBR6 focused on mountain rivers of the Alps and their surroundings, and speciﬁcally addressed the European challenge in terms of ecological improvement in a cultural landscape where natural hazards are critical. In the host country of Austria, the issue of natural hazards is of immediate relevance due to ﬂoods in August 2002 and August 2005 – the latter immediately prior to the conference – that had recurrence intervals of more than 1000 years and induced signiﬁcant morphological changes in many gravel-bed rivers, including threeto four-fold increases of the river bed width and metamorphosis from single thread to braided rivers. The implementation of the European Water Framework Directive strives to achieve a ‘‘good ecological status’’ of all running waters by the year 2015. River basin management is crucial in this respect because gravel-bed river behaviour is determined by a complex interaction of large- and small-scale processes and patterns. In general, Alpine rivers have undergone signiﬁcant changes over the last two centuries. Human activities have modiﬁed their geometry through engineering measures to gain land for agricultural purposes and settlements, as well as through active mining to exploit gravel resources. Their sediment and water transfers have also been altered by hydropower plant construction, torrent control works, and catchment land-use changes. The resulting river morphological changes have led to abiotic (e.g., river bed

vi

Foreword

degradation and narrowing) and biotic (e.g., longitudinal and lateral disconnection) disruption. In river basins in the Alps, where rivers are not steady-state but follow long-term trajectories of changes related to multiple human driven parameters, the current management situation has been made critical by channel instability problems, ﬂood effects and biodiversity decrease, and river restoration is a major issue. Early attempts at river restoration mainly focused on small-scale measures. Today, successful restoration projects in high-energy and bedload transport-dominated conditions must include the full spectrum of scales and initiate self-forming morphodynamics. Three sessions of GBR6 dealt with scales, ranging from ‘‘scales of analysis for gravel-bed rivers’’, ‘‘analysis of processes at point and local scale’’, and ‘‘basin scale: sediment delivery and storage’’. One session treated ‘‘channel change and instability’’, another session covered ‘‘ecohydrology and ecohydraulics’’ and the last session focussed on ‘‘management and restoration’’. Furthermore, an extended PhD session was held. Each session was designed to contain one geomorphologist, one engineer and one ecologist in order to promote interdisciplinary discussion and to stimulate future interrelations and collaborations across the ﬁelds. Traditionally, intensive ﬁeldwork is undertaken during GBR events, where practical questions dealing with gravel-bed river processes and management are exposed to the participants and discussed. During GBR6, several speciﬁc reaches of the Isel River, the Upper Drau River, and torrential tributaries were considered in terms of three themes: sediment sources and torrent control; sediment input, transfer and river management; and river restoration design based on hydromorphological trends. Tour guides provided additional information to the participants. The ﬁnal discussion of the ﬁeldwork was incorporated into a Carinthian BBQ that took place along the Drau River. This book presents invited papers that were at least double-blind reviewed, thus maintaining the high standard of earlier gravel-bed river books. The book is organized into six parts:

Introductive contributions deal with the scale of analysis, with particular focus on scaling processes by M. Church, reach characterisation by R. Ferguson, and hydrodynamics and turbulence by V. Nikora. The second part is devoted to the analysis of the river processes at the point and local scales, with disciplinary questions in hydraulics, physics and ﬂuid mechanics. Speciﬁc contributions open discussions on bedload transport by P. Diplas, surﬁcial and sub-surﬁcial velocity by M. Detert and his colleagues, and velocities between the stream and the hyporheic zone by I. Seydell and her colleagues. At local scales, themes include bifurcation in braided rivers by M. Tubino and W. Bertoldi, geomorphic effects of ﬂoods by E. Mosselman and K. Sloff and by E.Wohl, bank erosion modelling by M. Rinaldi and S. Darby, and grain size responses to hydrological patterns by G. Parker and his colleagues. The third part focuses on the basin scale, including sediment delivery and storage, with speciﬁc discussions on sediment delivery and climate change by T. Coulthard and his colleagues, sediment delivery and human controls by M. Page and his colleagues, catchment responses to human activities and climate change by J. Pizzuto and his colleagues, sediment transport and its link to sediment supply by S. Ryan and M. Dixon, sediment organisation at the basin scale by J. Hoyle and her

Foreword

vii

colleagues, evolution of sediment waves by T. Lisle, and sediment storage and transport in coarse-bed streams by M. Hassan and his colleagues. The fourth part introduces more applied questions related to channel change and instability. Speciﬁc contributions concern a review of ecological responses due to human pressures by F. Nakamura and his colleagues, channel incision by B. Wyz˙ ga, vegetation encroachment of braided rivers by M. Hicks and his colleagues, and the effects of extreme ﬂoods on channel processes and stability by M. Jaeggi. The ﬁfth part deals with ecohydrology and ecohydraulics, linking hydraulic and geomorphic processes with ecological demands and providing scientiﬁc knowledge for river managers. The chapters include reservoir operation and ecosystem losses by K. Jorde and his colleagues, hydraulic effects on macroinvertebrate communities at the local scale by S. Rice and his colleagues, hydraulic geometry and ecological implications by N. Lamouroux, and gravel bars as key habitats for vegetation by D. Gilvear and his colleagues. The sixth part concludes the book with clearly applied contributions to river management and restoration, providing a large set of gravel-bed river examples, with a review of restoration experiences in the Alps and their surroundings by H. Habersack and H. Pie´gay, a discussion on uncertainty in river restoration by D. Sear and his colleagues, the evolutionary scenario and its use for the development of a conservation and restoration strategy for the Willamette River by S. Gregory, and the ecological assessment of restoration on the Drau River, Austria by S. Muhar and her colleagues.

Following this peer-review book publication, submitted poster papers, after successful review, have also been published in special sister issues in Earth Surface Processes and Landforms and Geodinamica Acta. Special thanks go to Peter Ergenzinger and Trevor Hoey as members of the organizing committee, Hugo Seitz for coordinating the local workshop organisation, Fre´de´ric Lie´bault and Jacqueline Dupuis for their active participation in the scientiﬁc organisation, John Laronne for very valuable inputs during the genesis of the workshop and also his constant attention to guide us in the right direction with respect to the contributors, all the colleagues involved in the pre- and post-conference tours, the sponsors of the workshop, the Austrian Ministry of Agriculture, Forestry, Environment and Water Management, the regional water authorities of Tyrol and Carinthia, the village of St. Jakob/Defereggental, the company ‘‘Interconvention’’ for all administrative support, convenors of the pre- and post-study excursions, chairmen of the sessions and many students for helping to prepare and convene the workshop. Due to the excellent contributions and fruitful discussions, the workshop was of very high quality. The venue’s Austrian mountain scenery, combined with the smallvillage atmosphere, wonderful weather and social activities, will hopefully ﬁrmly implant GBR6 in the memories of the participants and entice some to return to explore these Alpine gravel-bed rivers and their environments in more detail. Helmut, Herve´, Massimo Developments in Earth Surface Processes

Invited Participants in the 6th Gravel-Bed Rivers Conference 2005

Contents Foreword List of contributing authors

v xiii

Scales of analysis for gravel-bed rivers 1 2 3

Multiple scales in rivers Michael Church Gravel-bed rivers at the reach scale Rob Ferguson Hydrodynamics of gravel-bed rivers: scale issues Vladimir Nikora

3 33 61

Analysis of processes at point and local scales 4

Pressure- and velocity-measurements above and within a porous gravel bed at the threshold of stability Martin Detert, Michael Klar, Thomas Wenka and Gerhard H. Jirka 5 Evaluating vertical velocities between the stream and the hyporheic zone from temperature data Ina Seydell, Ben E. Wawra and Ulrich C.E. Zanke 6 Bifurcations in gravel-bed streams Marco Tubino and Walter Bertoldi 7 The importance of ﬂoods for bed topography and bed sediment composition: numerical modelling of Rhine bifurcation at Pannerden Erik Mosselman and Kees Sloff 8 Review of effects of large ﬂoods in resistant-boundary channels Ellen Wohl 9 Modelling river-bank-erosion processes and mass failure mechanisms: progress towards fully coupled simulations Massimo Rinaldi and Stephen E. Darby 10 Adjustment of the bed surface size distribution of gravel-bed rivers in response to cycled hydrographs Gary Parker, Marwan A. Hassan and Peter Wilcock 11 Bed load transport and streambed structure in gravel streams Panos Diplas and Hafez Shaheen

85

109 133

161 181

213

241 291

Contents

x Basin scale: sediment delivery and storage 12

13

14

15

16

17

18

Non-stationarity of basin scale sediment delivery in response to climate change Tom J. Coulthard, John Lewin and Mark G. Macklin Changes in basin-scale sediment supply and transfer in a rapidly transformed New Zealand landscape Mike Page, Mike Marden, Mio Kasai, Basil Gomez, Dave Peacock, Harley Betts, Thomas Parkner, Tomomi Marutani and Noel Trustrum Two model scenarios illustrating the effects of land use and climate change on gravel riverbeds of suburban Maryland, U.S.A. Jim Pizzuto, Glenn Moglen, Margaret Palmer and Karen Nelson Spatial and temporal variability in stream sediment loads using examples from the Gros Ventre Range, Wyoming, USA Sandra E. Ryan and Mark K. Dixon Sediment organisation along the upper Hunter River, Australia: a multivariate statistical approach Joanna Hoyle, Gary Brierley, Andrew Brooks and Kirstie Fryirs The evolution of sediment waves inﬂuenced by varying transport capacity in heterogeneous rivers Thomas E. Lisle Sediment storage and transport in coarse bed streams: scale considerations Marwan A. Hassan, Bonnie J. Smith, Dan L. Hogan, David S. Luzi, Andre E. Zimmermann and Brett C. Eaton

315

337

359

387

409

443

473

Channel change and instability 19

20

21

22

Ecological responses to anthropogenic alterations of gravel-bed rivers in Japan, from ﬂoodplain river segments to the microhabitat scale: a review Futoshi Nakamura, Yoˆichi Kawaguchi, Daisuke Nakano and Hiroyuki Yamada A review on channel incision in the Polish Carpathian rivers during the 20th century Bart!omiej Wyz˙ga Contemporary morphological change in braided gravel-bed rivers: new developments from ﬁeld and laboratory studies, with particular reference to the inﬂuence of riparian vegetation D. Murray Hicks, Maurice J. Duncan, Stuart N. Lane, Michal Tal and Richard Westaway The ﬂoods of August 22–23, 2005, in Switzerland: some facts and challenges Martin Jaeggi

501

525

557

587

Contents

xi

Ecohydrology and ecohydraulics 23

24

25 26

Reservoir operations, physical processes, and ecosystem losses Klaus Jorde, Michael Burke, Nicholas Scheidt, Chris Welcker, Scott King and Carter Borden Movements of a macroinvertebrate (Potamophylax latipennis) across a gravel-bed substrate: effects of local hydraulics and micro-topography under increasing discharge Stephen P. Rice, Thomas Buffin-Be´langer, Jill Lancaster and Ian Reid Hydraulic geometry of stream reaches and ecological implications Nicolas Lamouroux Gravel bars: a key habitat of gravel-bed rivers for vegetation David Gilvear, Robert Francis, Nigel Willby and Angela Gurnell

607

637 661 677

River management and restoration 27

28 29

30

River restoration in the Alps and their surroundings: past experience and future challenges Helmut Habersack and Herve´ Pie´gay Uncertain restoration of gravel-bed rivers and the role of geomorphology David A. Sear, Joseph M. Wheaton and Stephen E. Darby Historical channel modiﬁcation and ﬂoodplain forest decline: implications for conservation and restoration of a large ﬂoodplain river – Willamette River, Oregon Stanley Gregory Restoring riverine landscapes at the Drau River: successes and deﬁcits in the context of ecological integrity Susanne Muhar, Mathias Jungwirth, Gu¨ nther Unfer, Christian Wiesner, Michaela Poppe, Stefan Schmutz, Severin Hohensinner and Helmut Habersack

Subject index

703 739

763

779

809

List of contributing authors Note: Bold names indicate the corresponding authors. Walter Bertoldi Dipartimento di Ingegneria Civile e Ambientale, University of Trento, Trento, Italy Harley Betts

Landcare Research, Palmerston North, New Zealand

Carter Borden

Center for Ecohydraulics Research, University of Idaho – Boise, USA

Gary Brierley School of Geography and Environmental Science, University of Auckland, Auckland, New Zealand Andrew Brooks Centre for Riverine Landscapes, Grifﬁth University, Nathan, Queensland, Australia Thomas Buffin-Be´langer Module de ge´ographie, De´partement de biologie, chimie et ge´ographie, Universite´ du Que´bec a` Rimouski, Canada Michael Burke

Center for Ecohydraulics Research, University of Idaho – Boise, USA

Michael Church Department of Geography, University of British Columbia, Vancouver, British Columbia, Canada, [email protected] Tom J. Coulthard Department of Geography, University of Hull, Hull, UK, [email protected] Stephen E. Darby

School of Geography, University of Southampton, Highﬁeld, UK

Martin Detert Institute for Hydromechanics (IfH), University of Karlsruhe, Germany, [email protected] Panos Diplas Department of Civil and Environmental Engineering, Virginia Tech, Blacksburg, USA, [email protected] Mark K. Dixon USA

USDA Forest Service, Rocky Mountain Research Station, Fraser,

Maurice J. Duncan

NIWA, Christchurch, New Zealand

List of contributing authors

xiv

Brett C. Eaton Department of Geography, The University of British Columbia, Vancouver, British Columbia, Canada Rob Ferguson Department of Geography, Durham University, Durham, UK, [email protected] Robert Francis London, UK

Department of Geography, Kings College London, Strand.

Kirstie Fryirs Department of Physical Geography, Macquarie University, North Ryde, NSW, Australia David Gilvear School of Biological and Environmental Sciences, University of Stirling, UK, [email protected] Basil Gomez Geomorphology Laboratory, Indiana State University, Terre Haute, Indiana, USA Stanley Gregory Department of Fisheries & Wildlife, Oregon State University, Corvallis, Oregon, USA, [email protected] Angela Gurnell

Department of Geography, Kings College London, London, UK

Helmut Habersack Institute of Water Management, Hydrology and Hydraulic Engineering, Department of Water, Atmosphere and Environment, BOKUUniversity of Natural Resources and Applied Life Sciences, Vienna, Austria, [email protected] Marwan A. Hassan Department of Geography, The University of British Columbia, Vancouver, British Columbia, Canada, [email protected] D. Murray Hicks

NIWA, Christchurch, New Zealand, [email protected]

Dan L. Hogan British Columbia Ministry of Forests, PO Box 9519, Station Provincial Government, Victoria, British Columbia, Canada Severin Hohensinner Institute of Hydrobiology & Aquatic Ecosystem Management, Department of Water, Atmosphere & Environment, BOKU – University of Natural Resources and Applied Life Sciences, Vienna, Austria Joanna Hoyle Department of Physical Geography, Macquarie University, North Ryde, NSW, Australia, [email protected] Martin Jaeggi River Engineering and Morphology, Ebmatingen, Switzerland, [email protected]

List of contributing authors Gerhard H. Jirka Germany

xv

Institute for Hydromechanics (IfH), University of Karlsruhe,

Klaus Jorde Center for Ecohydraulics Research, University of Idaho – Boise, USA, [email protected] Mathias Jungwirth Institute of Hydrobiology & Aquatic Ecosystem Management, Department of Water, Atmosphere & Environment, BOKU – University of Natural Resources and Applied Life Sciences, Vienna, Austria Mio Kasai Australia

Department of Physical Geography, Macquarie University, NSW,

Yoichi Kawaguchi Aqua Restoration Research Center, Public Works Research Institute, Gifu, Japan Scott King USA

Center for Ecohydraulics Research, University of Idaho – Boise,

Michael Klar Robert Bosch GmbH; formerly, Interdisciplinary Center for Scientiﬁc Computing (IWR), University of Heidelberg, Germany Nicolas Lamouroux CEMAGREF, UR Biologie des Ecosyste`mes Aquatiques, Lyon, France, [email protected] Jill Lancaster Institute of Evolutionary Biology, School of Biological Sciences, University of Edinburgh, Edinburgh, UK. Stuart N. Lane

Department of Geography, University of Durham, Durham, UK

John Lewin Institute of Geography and Earth Sciences, University of Wales, Aberystwyth, UK Thomas E. Lisle USDA Forest Service, Paciﬁc Southwest Research Station, Arcata, California, USA, [email protected] David S. Luzi Department of Geography, The University of British Columbia, Vancouver, British Columbia, Canada Mark G. Macklin Institute of Geography and Earth Sciences, University of Wales, Aberystwyth, UK Mike Marden

Landcare Research, Gisborne, New Zealand

Tomomi Marutani Japan

Graduate School of Agriculture, Hokkaido University, Sapporo,

List of contributing authors

xvi

Glenn Moglen Department of Civil and Environmental Engineering, University of Maryland, College Park, USA Erik Mosselman Delft University of Technology & WL|Delft Hydraulics, Delft, The Netherlands, [email protected] Susanne Muhar Institute of Hydrobiology & Aquatic Ecosystem Management, Department of Water, Atmosphere & Environment, BOKU – University of Natural Resources and Applied Life Sciences, Vienna, Austria, [email protected] Futoshi Nakamura Graduate School of Agriculture, Hokkaido University, Sapporo, Japan, [email protected] Daisuke Nakano Japan

Graduate School of Agriculture, Hokkaido University, Sapporo,

Karen Nelson Department of Entomology, University of Maryland, College Park, USA Vladimir Nikora Engineering Department, Kings College, University of Aberdeen, Aberdeen, UK, [email protected] Mike Page Institute of Geological and Nuclear Sciences, Lower Hutt, New Zealand, [email protected] Margaret Palmer

UMCES Chesapeake Biological Laboratory, Solomons, USA

Gary Parker Department of Civil and Environmental Engineering and Department of Geology, University of Illinois, Urbana, USA, [email protected] Thomas Parkner Graduate School of Agriculture, Hokkaido University, Sapporo, Japan Dave Peacock

Gisborne District Council, Gisborne, New Zealand

Herve´ Pie´gay University of Lyon, CNRS-UMR 5600 EVS, Site Ens-lsh, Lyon, France, [email protected] Jim Pizzuto Department of Geology University of Delaware, Newark, USA, [email protected] Michaela Poppe Institute of Hydrobiology & Aquatic Ecosystem Management, Department of Water, Atmosphere & Environment, BOKU – University of Natural Resources and Applied Life Sciences, Vienna, Austria Ian Reid

Department of Geography, Loughborough University, UK

List of contributing authors

xvii

Stephen P. Rice Department of Geography, Loughborough University, UK, [email protected] Massimo Rinaldi Dipartimento di Ingegneria Civile e Ambientale, Universita` di Firenze, Firenze, Italy, [email protected] Sandra E. Ryan USDA Forest Service, Rocky Mountain Research Station, Fort Collins, USA, [email protected] Nicholas Scheidt USA

Center for Ecohydraulics Research, University of Idaho – Boise,

Stefan Schmutz Institute of Hydrobiology & Aquatic Ecosystem Management, Department of Water, Atmosphere & Environment, BOKU – University of Natural Resources and Applied Life Sciences, Vienna, Austria David A. Sear soton.ac.uk

School of Geography, University of Southampton, UK, D.Sear@

Ina Seydell Institute of Hydraulic and Water Resources Engineering, University of Technology Darmstadt, Germany, [email protected] Hafez Shaheen

An-Najah National University, Palestine

Kees Sloff Delft University of Technology & WL Delft Hydraulics, Delft, The Netherlands Bonnie J. Smith Department of Geography, The University of British Columbia, Vancouver, British Columbia, Canada Michal Tal St Anthony Falls Laboratory, National Center for Earth-Surface Dynamics, University of Minnesota, Minneapolis, MN, USA Noel Trustrum Zealand

Institute of Geological and Nuclear Sciences, Lower Hutt, New

Marco Tubino Dipartimento di Ingegneria Civile e Ambientale, University of Trento, Trento, Italy, [email protected] Gu¨nther Unfer Institute of Hydrobiology & Aquatic Ecosystem Management, Department of Water, Atmosphere & Environment, BOKU – University of Natural Resources and Applied Life Sciences, Vienna, Austria Ben E. Wawra Institute of Hydraulic and Water Resources Engineering, University of Technology Darmstadt, Germany

List of contributing authors

xviii Chris Welcker USA

Center for Ecohydraulics Research, University of Idaho – Boise,

Thomas Wenka Federal Waterways Engineering and Research Institute (BAW), Karlsruhe, Germany Richard Westaway UK Joseph M. Wheaton

Halcrow Group Limited, Burderop Park, Swindon SN4 0QD,

School of Geography, University of Southampton, UK

Christian Wiesner Institute of Hydrobiology & Aquatic Ecosystem Management, Department of Water, Atmosphere & Environment, BOKU – University of Natural Resources and Applied Life Sciences, Vienna, Austria Peter Wilcock Department of Geography and Environmental Engineering, Johns Hopkins University, Baltimore, USA Nigel Willby Stirling, UK

School of Biological and Environmental Sciences, University of

Ellen Wohl Department of Geosciences, Colorado State University, Fort Collins, CO, USA, [email protected] Bart#omiej Wyz˙ga krakow.pl

Polish Academy of Sciences, Kracow, Poland, wyzga@iop.

Hiroyuki Yamada Japan

Graduate School of Agriculture, Hokkaido University, Sapporo,

Ulrich C.E. Zanke Institute of Hydraulic and Water Resources Engineering, University of Technology Darmstadt, Germany Andre E. Zimmermann Department of Geography, The University of British Columbia, Vancouver, British Columbia, Canada

Gravel-Bed Rivers VI: From Process Understanding to River Restoration H. Habersack, H. Pie´gay, M. Rinaldi, Editors r 2008 Elsevier B.V. All rights reserved.

3

1 Multiple scales in rivers Michael Church

Abstract Rivers are characterized by multiple scales of length and time. A viable theory of river behaviour must reconcile the various processes that occur at different scales. Scales of turbulent ﬂuid motion are determined by the properties of the ﬂuid, while the texture of boundary materials formally scales criteria for resistance to ﬂow and sediment entrainment. The macroscale limit of turbulent motion is the scale at which the ﬂow detects its conﬁning boundaries. The size of the channel is determined by how much water the channel must pass and is summarized by channel width. Channel scales also depend on the nature and magnitude of sediment ﬂuxes and the valley gradient down which they must be passed, while the evolution of channel morphology is directed by its interaction with the persistent secondary circulation. From these conditions a set of scales is derived that deﬁne the channel state. At scales between that of channel width and some limit scale set by the larger landscape, river channel pattern may exhibit scale-free (self-similar) behaviour while, at still larger scales, river systems are subject to topographical constraints set by valley form and bedrock structure. The latter conditions are mainly externally set and contingent. Superimposed on these physical scales is a set of ecological scales expressed by aquatic organisms. These are derivative insofar as life is evolutionary and adaptive but, in this area, there remains a great deal to learn.

1.

Introduction

Rivers are the product of turbulent stream ﬂow over myriad granular elements, vegetable matter, and solid refractory materials under strong, irregular forcing on effective timescales that range from seconds to centuries. They are complex systems because they are characterized by multiple scales of time and length that deﬁne a tightly integrated hierarchical set of subsystems, each of which exhibits some range of scale-free behaviour. A viable theory of river behaviour must reconcile processes that occur at different scales. It must also reconcile intrinsic dynamical scales – ones E-mail address: [email protected] ISSN: 0928-2025

DOI: 10.1016/S0928-2025(07)11111-1

Michael Church

4

associated with process – with extrinsic scales imposed by the materials that make up the river boundary. A successful analysis must demonstrate the consistency of ﬂuvial processes across scales whilst providing us with a satisfactory basis for understanding. Rivers surely represent one of the more difﬁcult problems in physical science. The purpose of this paper is to explore the hierarchy of scales that must be addressed in order to make progress toward understanding rivers and to examine some of the connections amongst them. These scales pose certain constraints to practical modes of analysis. Some constraints arise from the state of our technology, but others appear to be quite fundamental insofar as they are connected with the information that it appears possible to have about the river. The discussion will remain elementary: the study of scales in ﬂuvial phenomena has remained almost entirely empirical, and it appears most important to focus attention on what might be the fundamental scales of interest. First, we must establish what is a scale in a physical problem.

2.

Scales, scaling, and parameterization

The ‘‘scale’’ of a phenomenon is widely but informally recognized as the (order of) magnitude, in length and time, of the principal elements and/or processes involved. In particular problems, other dimensions – for example, mass – may also be relevant. By this deﬁnition, we recognize the scale of turbulent phenomena, for example, to fall approximately within the range 104–101 m, while stream channels fall characteristically within the range 100–103 m (in width). More formally, ‘‘scaling’’ a problem entails the selection of intrinsic reference quantities (scales) so that each term in the dimensional equations that describe the problem is transformed into the product of a constant factor that closely estimates the term’s order of magnitude and a dimensionless factor of order 1.0. An ‘‘intrinsic’’ reference quantity is one that is endogenous to the problem, that is, one that arises directly from the physical properties of the medium and is therefore universal. The length scale, n/u, in which n is the kinematic viscosity and u the shear velocity, represents an intrinsic reference quantity for problems associated with the small scales of turbulence – the limit of turbulent motions set by ﬂuid viscosity. The laminar sublayer, for example, is scaled as uys/n, where ys is the dimensioned sublayer thickness. It is increasingly recognized that many phenomena occupy some range of scales speciﬁed by separate upper and lower limits. Within that range, the phenomena may exhibit essentially scale-free or ‘‘scaling’’ behaviour. The identiﬁcation of ranges of self-similar behaviour has also come to be known as ‘‘scaling’’. Scaling behaviour has received much attention lately, but it seems that the more fundamental knowledge about a phenomenon remains with the deﬁned limit scales. It is not always obvious how to choose scales. Whilst it appears most appropriate to choose them in the fundamental dimensions of mass, length and time, it is often convenient to select derivative quantities, such as reference velocities or ﬂuxes. Very often, problems in ﬂuvial hydraulics are not completely speciﬁed by known equations, and dimensional analysis is used to make a formal selection of appropriate

Multiple scales in rivers

5

parameter groups that include suitable scaling quantities. Dimensional analysis guarantees the requirement that properly scaled terms be rational (dimensionally balanced). ‘‘Parameterization’’ is the selection of a convenient (i.e., easily measured) surrogate variate that tracks the behaviour of a more fundamental variate of interest. The substitution is commonly effected via approximate theoretical arguments or by study of empirical correlations. Such procedures may be adopted to avoid the complications of analyzing complex subsystems. It is possible that such procedures can be used to discover appropriate scaling quantities. In this paper, most of the scales are descriptive and geometrical ones since we are seeking, in the main, to identify characteristic length and time scales of ﬂuvial processes. They have mostly been identiﬁed by empirical or formal means rather than from rational theory.

3.

The scales of turbulent ﬂow

The classical construction of turbulent ﬂow holds that the phenomenon is essentially random and the description of it, necessarily statistical. Two important length scales have been associated with this view, a turbulent microscale – the limit scale for turbulent motion at which viscous dissipation of kinetic energy becomes dominant – and an integral length scale, or macroscale, the largest scale at which spatial correlation persists in the ﬂow. A microscale pertinent to ﬂuvial processes is n/u, noted above. (This quantity corresponds with neither of the usually quoted energetic turbulent microscales.) The usual point velocity measurement technique R k leads to an approximation of the macroscale via the integral timescale T E ¼ o RðtÞdt, where R(t) denotes the correlation of velocity (or of some other property of the ﬂow) at increasing temporal intervals, t, k is deﬁned by the condition R(k)E0, and subscript E an Eulerian reference frame. Adopting the ‘‘frozen turbulence’’ convention (Taylor, 1935), so that a space-for-time substitution may be effected, LE ¼ ou4 TE then gives the macroscale, where ou4 is the temporal mean ﬂow velocity. This scale has been interpreted as the maximum dimension of a turbulent eddy in the direction of the measurement (usually downstream). Unlike the microscale, it betrays no overt clue about the reason for its existence. In the 1950s, ideas began to circulate about large-scale structure in supposedly random turbulence. In the west, early work consisted of attempts to interpret the macroscale in physical terms, and it was motivated by the desire to understand the compound form of turbulent velocity proﬁles (for a review of early work see Cantwell, 1981). On the basis of correlation studies and visualization work, Townsend (1970, 1976) proposed relatively early on that large-scale turbulent structures took the form of inclined contra-rotating cylinders of circulating water, or inclined cones, that extended through the water column. The latter conception acknowledges that structures originating in the near-bed region of high shear might grow in size as they move away from the boundary. Velikanov (1949) proposed a sequence of depth-scaled roller eddies. Later Russian work (Dement’ev, 1962; Makaveev, 1964; see Shvidchenko and Pender, 2001, for a brief summary of

6

Michael Church

additional work) apparently conﬁrmed this interpretation, but found that the features were streamwise elongated. American experimental work focused on the near-wall region where turbulent kinetic energy production was found to be concentrated (Klebanoff, 1954) and led to the ‘‘Stanford model’’ (Kline et al., 1967; Corino and Brodkey, 1969) of bursting streaks. The model of uplifted (‘‘ejected’’), paired streaks of low-velocity water, which are replaced by an inrushing ‘‘sweep’’ of high-velocity water, was initially supposed to control the net production of turbulent energy in shear ﬂow and to be a wall-associated phenomenon governed by the turbulent microscale. It was therefore a major shock when Rao et al. (1971) reported that the burst period scales with outer region variates uN and d, the free stream velocity and boundary layer thickness, with the mean dimensionless time between bursts being given by uNT/dE5. Moreover, this scaling appeared to be characteristic throughout the depth of the ﬂow and the large eddies appeared to be advected at a rate comparable with the mean ﬂow, so their spacing and, conceivably, their size should be approximately LE5d. Grass (1971), in a critical contribution that compared turbulence characteristics over smooth and rough walls (the latter comprising 9 mm pebbles), showed that ejections and inrushes were present regardless of surface roughness. The near-wall ﬂow over a fully rough boundary must be substantially different from that over a smooth wall, yet the basic organization of the ﬂuid motion appears to remain largely unchanged. One is forced to the conclusion that the bursting process occurs in turbulent shear ﬂow irrespective of the boundary condition. One speculates that the reservoir of slow-moving ﬂuid resident in the laminar sublayer of smooth-wall ﬂows is replaced by reservoirs of low-momentum ﬂuid trapped amongst the elements of rough walls (Kirkbride, 1993). So the burst model has become a standard representation for momentum transfer and structure throughout turbulent shear ﬂows (see, e.g., Jackson, 1976; Yalin, 1992; Best, 1993), with the implication that the wall-based phenomena associated with turbulent energy production control the character of the entire turbulent ﬂow. This appearance raises the questions whether the scaling of Rao et al. represents a suitable turbulent macroscale, and what is its relation to LE. Subsequent experimental studies and initial ﬁeld studies have yielded additional estimates of the scale of coherent structures, a summary of which is given in Table 1.1. The results must be appraised with considerable reservation. First, d is ﬂow depth, d, not boundary layer depth. Since most rivers have truncated boundary layers (i.e., depth is not sufﬁcient for the boundary layer to fully develop), the true value of d remains unknown. Similarly, u mostly is U, the proﬁle averaged mean velocity of the ﬂow, rather than uN. The effects of these two biases tend to mutually cancel, but the underestimate of depth likely is generally the greater so that computed values may be fractionally high. Most of the values listed for L/d fall in the range 2–7, which is comparable with the range 3–7 that has been reported (Jackson, 1976). A more fundamental problem is that turbulent bursts and large eddies actually occur intermittently and the distribution of their recurrence period is skewed (see Fig. 1.1) – in fact, exhibits a scaling range. So it is not clear, on the analysis of the complete velocity record, just what may be true ejection events, hence just what is the actual frequency of the burst cycle. Investigations might fail to detect certain events – inﬂating the time and length scales – or may mix various derivative eddies with the bursts, producing negatively biased estimates.

Multiple scales in rivers Table 1.1.

Data of macroturbulent eddy scales. ua

db

Dc

L/dd

Remarks

0.07 0.73 0.38–0.98

0.04 0.12 0.03–0.10

– 4.8 2–8

2.1 3.0 4.770.2

0.3 m wide ﬂume 0.6 m wide ﬂume 0.3 m wide; n ¼ 18

0.36 0.34

0.35–0.40 0.28

33 30–50

6.7–6.9 9.6

Gravel-bed river Gravel-bed river

0.45–0.67

0.35–0.60

30–45

2–6e

Gravel-bed river

1.5–1.7

1.7–2.1

6.470.1

Jackson (1976)

0.64–1.42

0.55–5.6

7.871.5

Kostaschuk et al. (1991)

1.0

11

3.9–4.6

n ¼ 3; 1 datum deleted 0.2 m dunes; n ¼ 14 2 m dunes

Source Flume studies Komori et al. (1982) Cellino and Graf (1999) Shvidchenko and Pender (2001) Gravel-bed rivers Bufﬁn-Be´langer et al. (2000) Paiement-Paradis et al. (2003) Roy et al. (2004) Sand-bed rivers Korchokha (1968)

a

7

1

Values in m s ; should be uN, but most values are mean velocities, Jackson and Kostaschuk values are surface velocities. b Values in m; all values are ﬂow depths. c Values in mm; values are D50 of mixture or size of homogeneous material. d Calculated by writer as uT/d. T is time between successive observed structures or mean duration for passage of a structure. e Full range is 1–10 due to two excentric values. Many determinations were made from data collected by three independent methods. See source for details.

The scaling of Rao et al. is in fact very closely allied with the Strouhal number, the scaled interval for eddy shedding from a bluff object (in which the classical constant is 2p, but the ﬁgure is lower at high Reynolds number). Eddies certainly are shed from individual protruding objects on rough beds, and so it appears quite possible that the apparently universal phenomenon of macroturbulent bursting represents a general phenomenon of eddy production over a rough boundary – the consequence of a sort of ‘stick and slip’ motion of the ﬂuid over the surface. Jackson (1976) investigated the relation between evident bursting frequencies and the Eulerian macroscale using the turbulence measurements collated by McQuivey (1973) and concluded that they are essentially equivalent. On that evidence, it appears that L ¼ uNT might represent a suitable turbulent macroscale. A deeper issue is presented by the appearance, in Fig. 1.1, that the occurrence of turbulent eddies is scale-free over the range between the scale of their generation near the bed and some multiple of channel depth. This reinforces the notion that channel depth functions in some way as a meaningful limit scale for turbulent eddies. (Most studies of macroturbulent structure to date have been conducted in essentially 2-D ﬂows; on the other hand, available information on the eddies’ lateral structure also implicates d as a suitable unit scale.) The balance of evidence suggests that 3rL/ dr6, perhaps, but it is possible that the upper limit value should be preferred, which would tend to conﬁrm or magnify the original estimate of Rao et al. (1971), and this would no longer be a ‘‘characteristic’’ scale, but the limit of a scaling range.

Michael Church

8 103

NUMBER OF EVENTS

102

101

100

10-1 -2 10

10-1

100

101

DURATION (S) Figure 1.1. Frequency distribution of the duration of burst-type events in a shallow ﬂow in a gravel-bed river over 30–45 mm cobbles (R. Eaton Nord, Que´bec, Canada; data from Roy et al., 2004). Closed symbols indicate the record of all events initiated by an upward passage of the velocity past ou4; open symbols indicate the same record restricted to events initiated when velocity exceeds ou4+1.3su, the standard deviation of velocity. The plot can be scaled to eddy length by multiplying by the mean velocity, 0.62 m s1. At the low-frequency end, the two plots merge: it is most likely that eddies at this scale reﬂect the burst period. (Reproduced with permission from Cambridge University Press.) Similar data have been presented by Jackson (1976) and by Lapointe (1992) for ﬂows in sand-bed rivers with dunes. The plots always exhibit signiﬁcant ranges of self-similarity.

4.

Roughness scales

Turbulence production in river channels is complicated by the fact that river ﬂows are shear ﬂows over a more or less rough boundary. In gravel-bed rivers, the boundary certainly is rough; that is, the individual clastic elements forming the boundary represent a signiﬁcant source of resistance to the ﬂow. This fact is important when one considers the application of theoretical and experimental results on turbulence to ﬂows over gravels. Most experimental studies of turbulence have been conducted over smooth boundaries because the chief motivation has been to

Multiple scales in rivers

9

understand ﬂows over surfaces such as airfoils, turbine blades, and other machinery. The capacity of the rough boundary to generate eddies directly at scales determined by the scale of the boundary elements creates circumstances that may be signiﬁcantly different from those investigated in most ﬂuid mechanical experiments. In the region immediately above a rough surface, the mean proﬁle of turbulent shear ﬂow has been found to take the form u 1 30y (1.1) ¼ ln un k ks pﬃﬃﬃﬃﬃﬃﬃﬃ in which kE0.41 is a constant (von Ka´rma´n’s constant), un t=r a velocity scale (the shear velocity), and ks a roughness length scale. This equation is Prandtl’s ‘‘law of the wall’’ and shows that velocity is proportional to roughness-scaled distance (y) from the wall. When D/d-1.0, D being a characteristic bed material grain diameter, the description breaks down as the ﬂow becomes a series of jets between the large roughness elements. Conversely, when the ﬂow becomes deep, an outer layer develops above the wall layer. The logarithmic wall layer is often asserted to apply to the lower 15–20% of the ﬂow but, over high roughness, logarithmic dependence often extends right to the surface, but with modiﬁed scaling (Fig. 1.2). The division of the ﬂow proﬁle is related to the phenomena of turbulent eddy production and dispersion in the ﬂow. Production is concentrated at the base of the wall layer, but eddies remain highly coherent within the wall layer before dispersing upward, giving rise to a near-uniform distribution of ﬂuid shear within the wall layer. Near the bed, the ﬂow ‘‘sees’’ only the eddies produced locally but, higher up, the ﬂow is inﬂuenced by advected and dispersing eddies generated from a more extended upstream area. The roughness of the bounding material of the channel has classically been scaled by reference to bed material grain size. The roughness length, ks, can be related to a zero-plane displacement, y0 – the distance above the entirely solid bounding plane at which the velocity becomes zero. If we scale distance by y0, we have y0 ¼ ks/30. On granular boundaries, both ks and y0 are difﬁcult to deﬁne physically so that, conventionally, ks is parameterized as kD, k being a constant of proportion or ‘‘grain size multiplier’’. Gravel beds consist of a spectrum of grain sizes, the larger of which protrude signiﬁcantly into the ﬂow, so it is usual to adopt some large grain size (e.g., D84; D90) as the appropriate surrogate scale. This long-adopted practice is supported by the observation that the large grains support a high proportion of the entire shear stress exerted on the channel bed. van Rijn (1982) found from literature review that, for D90, the range of values assigned to k varies from 1 to 10. Clifford et al. (1992) demonstrated from measurements that the commonly quoted value ks ¼ 3.5D84 yields reasonable results over a relatively featureless gravel bed. However, no single constant is apt to be robust, since the effective roughness depends on not only grain protrusion but also the net effect of aggregate grain structures, so there is no welldeﬁned single scale. k is, in effect, a scale adjustment to cover a number of complexities of the boundary. Integration of equation (1.1) yields U 1 11d (1.2a) ¼ ln un k ks

Michael Church

10 10 surface

5.0

y - ks (cm)

2.0

1.0

1.9 cm above bed

0.5

1/20

1/16

0.2

1/8 1/80 1/12 0.1 10

20

30

40

U (cm/s) Figure 1.2. Velocity proﬁles over channel beds with high roughness (after Nowell and Church, 1979): experimental data using regularly arrayed roughness elements of 0.9 cm height and variable spacing. Fractions given for the individual proﬁles represent roughness density. In this representation, roughness length was assumed to be 0.9 cm. The proﬁles clearly show two segments, the lower of which corresponds with a zone of uniform shear for roughness densities41/48. Arbitrary adjustment of the roughness length can straighten the proﬁles, which yields a single velocity scale, but there is no physical reason to perform such an adjustment. The alternative is to accept distinct velocity scales, u, for the wall region and the outer region.

Multiple scales in rivers

11

and this result was further speciﬁed by Keulegan (1938) for a trapezoidal channel of ﬁnite width as U 1 12:2R (1.2b) ¼ ln un k ks in which R is the hydraulic radius. This equation speciﬁes the resistance to ﬂow over the rough boundary by specifying the scaled mean velocity, again in relation to the roughness length scale. The roughness scale is closely allied with the relative roughness, R/ksd/D, which is a somewhat special quantity insofar as it forms the ratio of two important system scales. In gravel-bed rivers, grain size appears to scale signiﬁcant bed structures that further modify the resistance to ﬂow. Most of the time, bed material moves in gravelbed rivers in a regime of size-selective partial transport (Wilcock and McArdell, 1993). The ﬁner materials are preferentially transported in comparison with the larger materials. The largest grains of all may rarely be entrained. Those large clasts become keystones for imbricate accumulation of grain clusters (Brayshaw, 1984) which may ramify into lines (Laronne and Carson, 1976) and irregular reticulate networks (Church et al., 1998; from which the following description is taken). Most of the clasts that form such structures are larger than D84 of the bed material, and the keystones might correspond with some size of order D90 or larger. The ratio of structure spacing (diameter) to constituent clast diameter is of order 10:1, implying that the constituent stones occupy between 15% (linear features) and 25% (stonebound circles) of the bed. These values correspond with the fractional area (approximately 0.12oao0.25) indicated by Rouse (1965) from analysis of experimental results (and observed in the data of Fig. 1.2) to contribute most of the boundary frictional resistance to ﬂow. These ﬁgures represent the range of concentrations for dominant roughness elements to produce intense wake interactions and strong turbulence production (Nowell and Church, 1979). The features appear to play a role in gravel-bed channels similar to primary bedforms in sand-bed channels. However, the latter appear to be scaled by d or d: the basis for scaling these emergent channel bed features probably is quite different in the two cases. An important distinction between them is sediment transport intensity – in the case of primary sandy bedforms, full bed mobility is achieved. In unusually powerful ﬂows, full mobility is achieved in gravel-bed channels, and then similar, depth-scaled bedforms develop, but usually with only low amplitude relative to wavelength. Roughness scales have characteristically been estimated in terms of an equivalent grain size, D, since this is the most obvious length scale associated with the rough boundary. It is evident, however, that the real complexity of the boundary prevents any single, unequivocally identiﬁable scale from emerging. In light of this, some investigators have adopted a spectral approach to determining roughness scales (early work is summarized in Nikora et al., 1998). It has emerged that the standard deviation of bed elevation – in effect, the vertical dimension of bed roughness – is a robust length scale for homogeneously rough beds (Aberle and Nikora, 2006) and is a viable scale for estimating ﬂow resistance even when D/d-1.0 (Smart et al., 2002; Aberle and Smart, 2003). But a signiﬁcant portion of ﬂow resistance – parameterized heretofore only through the use of the grain size multiplier – is associated with large

Michael Church

12

bedforms and with the channel geometry itself, so the problem of determining resistance to ﬂow is, of itself, a problem with multiple scales.

5.

Scaling sediment transport

The signiﬁcant fraction of stream sediment, for purposes of understanding alluvial channel form, is the bed material, the sediment that forms the bed and lower banks of the channel. Movement and deposition of this material alters the form of the channel. In gravel-bed channels, the displacement of bed material essentially corresponds with the movement of bedload. There is still no wholly rational theory from which to predict bed material or bedload movement. Since we require rational statements in order to study scaling, we follow many prior analysts in resorting to dimensional analysis. We consider the relevant variates r, the ﬂuid density; rs, the sediment density; n, ﬂuid viscosity; D, sediment particle diameter; d, water depth; t, the shear force per unit area imposed by the ﬂow on the bed; g, the acceleration of gravity; and gb, the mass sediment transport rate per unit channel width. These variates yield ﬁve dimensionless groups: Re ¼ uD/n, the grain Reynolds number; t ¼ t/gr(s1)D, the Shields number; c ¼ gb/g1/2r(s1)D3/2, Einstein’s transport intensity; s ¼ rs/r, the sediment speciﬁc weight; and D/d, the relative roughness. s is effectively a material constant, and it is known from experiment that Re does not vary signiﬁcantly within the range of gravel sizes. Hence, we have D (1.3) c ¼ f tn ; d t is, in effect, shear stress scaled according to D via the submerged particle weight, sediment speciﬁc weight remaining constant, which expresses the grain inertia. The important scale is, accordingly, the sediment grain scale (a reasonable result, but it emerges from the essentially arbitrary choice of D as a repeating variate in the dimensional analysis). In a fully alluvial channel, D constitutes an intrinsic scale of the system although, in many cobble- and boulder-bed channels, grain size of bed material is extrinsically imposed by material delivered from overbank by mass wasting. There exist a signiﬁcant number of speciﬁc realizations of equation (1.3), in most of which the term in relative roughness is, at best, implicit. An important part of specifying this equation in bulk calculations is selecting the appropriate grain scale. The values usually chosen to scale t have been the central measures Dm and D50, mean and median size respectively, the latter being preferred perhaps because it is not so sensitive to the variable skewness of grain size distributions. It has been shown (Wilcock and Southard, 1989) that, for purposes of estimating a bulk measure of bedload transport, D50 fairly represents the texture of sediment mixtures. An alternative approach to estimating sediment transport, which has the merit of relating the phenomenon directly to the deformation of the channel, is to consider the distance of transport and the volume (or mass) of sediment taking part in the process.

Multiple scales in rivers A simple, empirically scaled characterization of this process is u d s s C ¼ ðs 1Þð1 pÞ ou4 D

13

(1.4)

in which C is sediment fractional volumetric concentration, us ¼ xp/t the sediment virtual velocity (i.e., the distance traveled, xp, per unit time, including rest periods), ds the depth of the active layer taking part in the transport, and p the bed porosity but can also be adjusted to account for the reduced spatial density of partial transport. The scales ou4 and D are both accessible. ds is not easily measured, but it is possible that it is approximately constant at dsED90 (Wilcock and McArdell, 1997). This transport scaling has not been investigated. The characteristic step length of bed material in transport has received far less attention than it deserves. The ﬁrst systematic investigation was by Einstein (1937), who regarded particle displacement as a random process and conducted experiments to show that distributions of step lengths followed a gamma-related probability distribution. Tracing grains in natural streambeds is tedious and difﬁcult, and it is only in recent years that a number of observing programs have returned results that can be interpreted to reasonably reﬂect the displacement of the material. Pyrce and Ashmore (2003a) summarized results and have shown (Pyrce and Ashmore, 2003b) that, under weak transport, the distribution of displacements is, indeed, local and stochastic, as Einstein supposed but, under stronger transport, grains cluster and the characteristic distance correspond with the pool-rifﬂe spacing (Fig. 1.3) – that is, with the channel scale of transverse oscillation (discussed in the following section). This much makes elementary sense; bars are nothing more than aggregations of episodically mobile sediments. Nikora et al. (2002) have analyzed grain paths in more detail. Borrowing ideas from nonlinear statistical mechanics, they have deﬁned local, intermediate, and global ranges of displacement. Local displacement refers to a single trajectory between two points of collision with the bed, whilst an intermediate displacement refers to a sum of local displacements without intervening rest. Global displacements describe the sum of more than one intermediate displacement with intervening rest periods. They presented a dimensional analysis of the problem that issued in the formal relation xp D tun ¼ f tn ; Ren ; ; (1.5) ks D D wherein, oxp4 is the mean distance traveled by a particle, t the travel time (including rest periods for the global range), and the balance of the terms are earlier deﬁned. Ignoring Re, as usual, and supposing that ksED, or some constant proportion of it, as before, we obtain oxp4/D ¼ f[t, tu/D] with scale D, as one would expect. (Nikora et al. were actually interested in particle diffusion, hence in the higher moments of the particle position, but the scaling is similar.) Available data are restricted to experiments at ﬁxed ﬂows, hence ﬁxed values of t, and ﬁxed observing period, T, so that oxp4/D ¼ f[tu/D]. This scaling has not been investigated, but the results of Pyrce and Ashmore (2003b) indicate that there are ranges of behaviour in a manner analogous to those deﬁned by Nikora et al. in the time domain.

Michael Church

14 0.15 a) very weak transport (n = 160)

0.10

RELATIVE AMOUNT OF TRACER MATERIAL

0.05

0.00 0.15 b) intermediate transport (n = 1264)

0.10

0.05

0.00 0.15 c) strongest transport (n = 670)

0.10

0.05

0.00 0.00

0.25

0.50

0.75

1.00

1.25

1.50

DISTANCE DOWNSTREAM, (λ;meander wavelength) Figure 1.3. Distributions of bed material displacements, scaled by l, under variable ﬂow strength (ideal functions after Pyrce and Ashmore, 2003b; n ¼ number of experimental observations upon which the displayed distribution is based).

To this point, no distinction has been made amongst different grain sizes. Nikora et al. (2002) noticed that particle velocities do not vary systematically, at least over a restricted range of sizes, but systematic differences in travel distance have been demonstrated from ﬁeld data (presumably for intermediate and global ranges) amongst grains that initially were unconstrained (Church and Hassan, 1992). GrainsoD50 tend to show only moderate variation in travel distance, but travel distance of larger grains declines rapidly with size. This behaviour likely reﬂects the effect of the scaled size D/D50, smaller grains being liable to be trapped by larger ones resident on the bed, whereas the travel of larger grains is limited mainly by their inertia in relation to ﬂow strength. One supposes, then, that modulation of travel

Multiple scales in rivers

15

distance by grain size is a phenomenon of the partial transport regime. Church and Hassan showed that, as one would expect, the effects become stronger for grains that are restrained by clast structures.

6.

Channel scale

Once the ﬂow detects the boundary, it is steered by the boundary and, in turn, it may shape the boundary by erosion, transfer and deposition of the sediments of which it is formed. This gives rise to ‘‘pinned’’, therefore persistent, secondary circulations. These circulations scale with the channel dimensions. Channel width is scaled by the ﬂow that the channel must transmit, that is, by hydrology, and by bank materials. Therefore it is at least in part an extrinsic scale imposed on the ﬂuvial system. (By an ‘‘extrinsic scale’’, I mean one established by processes – such as watershed hydrology – that do not establish a universal scale for the ﬂuvial processes.) Water discharge represents one of the principal governing conditions imposed on river channels, the others being the quantity and caliber of sediment supplied to the channel from the drainage basin and the topographic gradient down which the ﬂuxes of water and sediment must be passed. Together, these conditions establish the hydraulic geometry of the channel, including the depth and mean velocity. Fundamental connections exist between the channel scale and the scales of turbulent ﬂow since the ﬂow and channel gradient (speciﬁcally, the product rgQS, in which Q is the water discharge, and S the energy gradient of the stream) establish the rate at which potential energy must be transformed into kinetic energy and dissipated in the turbulent ﬂow. Channel depth is arguably a more fundamental scale than width but it is not an intrinsic scale since the ratio of width to depth (the aspect ratio, w/d) again depends on material properties, in this case the bulk strength of the channel boundaries. Depth, however, appears to control the turbulent macroscale. Channel scale dynamics, in particular the phenomena associated with secondary circulations, are arguably the least well-understood aspect of ﬂuvial processes, probably because measurements remain logistically demanding, hence difﬁcult. The control of channel size by the ﬂow is recognized in the well-established regime relation w p Q1/2 (Leopold and Maddock, 1953), which is easily the most consistent of the oft-quoted equations of hydraulic geometry. This suggests w as an extrinsic channel scale, and the equation has been called a ‘channel scale relation’. It is, indeed, an empirical scaling relation, but an incomplete correlation. The complete set of independent governing conditions for river channel form informs the hydraulic geometry. The strength of the bed and bank materials is independently effective principally through bank materials, which may be affected by cohesion, cementation, or vegetable material. A rational theory of river regime must encompass all of these factors. Approaches to river regime have been essayed by many authors (see Ackers, 1988; Huang and Nanson, 2000; Grifﬁths, 2003; Eaton et al., 2004, for varying recent approaches). In the present context, the analysis by Eaton et al. is interesting since results are presented in a general, scaled form. They solved available equations specifying ﬂow continuity, ﬂow resistance, sediment transport, and bank strength

Michael Church

16

subject to the condition that ﬂow resistance be maximized and obtained a description of alluvial channel state in the scaled variates w/d, D/d, and t (Fig. 1.4). The length scales evidently are d and D. (It should be remarked that Eaton et al. made speciﬁc choices for parameterizing ﬂow resistance, sediment transport, and bank strength. The choices were guided by a desire to characterize gravel-bed channels, but the form of the solution space is unlikely to depend speciﬁcally on those choices.) In the theory constructed by Eaton et al., equations of hydraulic geometry for given boundary materials (i.e., on a plane of constant bank strength) are approximately functions of Q, S, and D. The appearance of D reﬂects the dependence of channel form, within the constraints set by the governing conditions, on processes worked out at the scales of turbulent ﬂow and sediment transport, which determine the channel dimensions within which the energy of the water is dissipated whilst passing the imposed sediment load. This dependence works through the total resistance to ﬂow. There is no general analysis for form resistance and the usual means of parameterizing it has been to adopt a multiplier for grain size, as discussed in the last section. Flow that is steered by the boundary gives rise to ‘‘pinned’’ eddies. These are the persistent secondary circulations that have been classically described (see review in Leliavsky, 1956) but are still not well analyzed. Eddies with transverse axes are potentially limited in their growth to order d, while those with vertical axes might grow to order w. Initiated by velocity gradients near the banks that give rise to pressure gradients within the ﬂow, the combination of these motions gives rise to

0.25 φ' = 60o

0.20 0.15

φ' = 50o

* 0.10

φ' = 40o

0.05 1

320 0.1

220

D/d 120 0.01

W/d

20

Figure 1.4. Alluvial state diagram (Eaton et al., 2004). j0 represents bank strength as measured by the effective friction angle. Each point on a plane of constant j0 represents an alluvial state characterized by unique Froude number, gradient (S), and sediment concentration (Qb/Q). The roles of t and S are equivalent and they may substitute for each other (reproduced with permission from John Wiley and Sons).

Multiple scales in rivers

17

helical circulations of channel scale with downstream axial orientation. These features may persist for some distance along the channel whilst having a subordinate dimension of order d, but their streamwise scale has not been independently established. They are delimited by channel geometry, but they simultaneously direct the currents that modify the geometry. The visually dominant form of deformation in all channels with erodible boundaries is lateral oscillation, whether it is expressed by a pool-rifﬂe sequence (Thompson, 1986) with wavelength l/2 or by more or less regular meanders with wavelength l. Superﬁcially, oscillatory behaviour, in the range 5wolo15w (Leopold et al., 1964), scales reasonably with channel width, which means that it can also be related to discharge. However, the origin of this scaling – indeed the question whether it is an appropriate scaling – remains obscure. The origin must be hydrodynamical since the homologous scale is present even in purely erosional channels in ice or bedrock (Leopold and Wolman, 1960) where, furthermore, the phenomenon can exhibit a 3-D, ‘‘corkscrew’’ character. But, since alluvial channel deformation entails sediment mobilization and redistribution, the scale in such channels must in some way be associated with sediment transfer, as demonstrated by the analyses of Pyrce and Ashmore. The scaling does not, evidently, arise directly from turbulent scales. Yalin (1977) proposed the existence of width-scaled eddies (or secondary ﬂows) originating from bank drag, by analogy with bed-generated eddies, for which a Strouhal-type shedding frequency might be appropriate, but the idea has not been much pursued (see, however, Clifford, 1993). Lateral diffusion of communicated information – in this case, represented by bed material transfer – may provide a physical basis to understand the expression of channel deformation. The most rapid pﬃﬃﬃﬃﬃﬃ communication of information in the ﬂow occurs as a gravity wave with vg ¼ gd , where vg denotes a cross-channel wave velocity. Suppose (see Eaton et al., 2006) that the net lateral spread of sediments proceeds at the rate vp ¼ avg, 0oao1. Then the material propagates across the channel with angle tan1(vp/oa0 u4) ¼ tan1(a00 /Fr), where Fr is the Froude number, and a0 , a00 are also fractional constants. If material originates in a scouring zone near one bank, and spreads to the opposite bank after a distance l/4 (realizing that the pool-rifﬂe spacing is l/2 and that entrained material will, on average, move half that distance), then, from geometry, w/(l/4) ¼ a00 /Fr and l ¼ (4Fr/a00 )w. This result may be compared with that from a linear stability analysis presented by Parker and Anderson (1975) and modiﬁed by Ikeda (1984) which pﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ gave l ¼ (5Fr/B)w, in which B ¼ Sðw=dÞ. These speculations provide a possible approach to investigate whether bed material diffusion is directly associated with the establishment of rifﬂe scaling distance, but there are, at present, few data with which to pursue the matter. Investigating the ratio l/w ¼ 4Fr/a00 (which conﬁnes attention to simple channels), limited data of Lewin (1976) and of Eaton and Church (2004) suggest that, for FrE0.85, aE0.5, l/ wE6.8, while experiments by Garcia and Nino (1993) in which l/wE9 suggest that a00 varies with Fr. These results are consistent with empirical experience on planform oscillation of rivers. Nevertheless, the arguments remain essentially formal and the underlying assumption is not established. A special case arises when D/d-1, so that individual large sediment grains or aggregations of a few grains are major, semi-stable form elements of the channel

Michael Church

18

boundary. This situation occurs most frequently on relatively steep gradients where large clasts are introduced into the channel by mass wasting from the adjacent valleyside slopes. The large grains control most or whole of the drop in the water surface, creating step-pool features with a natural scaling LED/S, in which L is steppool length and both D and S are imposed.

7.

Channel pattern

Channel units occur repeatedly in homogeneous reaches and exhibit elements of scale-free behaviour within speciﬁable limits. The evolution of channel form depends strongly on prior conﬁguration through its inﬂuence on sediment-directing circulation, so that an element of contingency enters into the analysis of a particular channel and reach-length river channel geometry does not exhibit a single, simple scale. This fact has been recognized in freely meandered rivers for a long time (Speight, 1965). Introduction of the concept of fractal geometry (Mandelbrot, 1977) opened a novel means to analyze scale-transcending geometry. Both the superposed sequence of bends in single-thread channels (Snow, 1989) and the nested sequence of bars in braided channels (Sapozhnikov and Foufoula-Georgiou, 1996) exhibit fractal ranges of dimension. But whereas mathematical fractals have no deﬁned scale limits, the scale-free (self-similar or self-afﬁne) behaviour of real objects has distinct limits imposed by material or energy constraints, or by larger scale features. The limits of selfsimilar scaling ranges may be more meaningful than the fractal character itself since these may suggest signiﬁcant limit scales for landscape processes (and certainly for modeling them). Nikora (1991) suggested that channel and valley width deﬁne the lower and upper limits of fractal ranges for river planform geometry. The lower limit is self-evident; the upper presumably deﬁnes a limit beyond which the planform is constrained in its mean direction of development hence, while larger structures might remain selfafﬁne, they could not remain self-similar. Beauvais and Montgomery (1996) found that the largest meander wavelength present bounds the possible self-similar scaling domain; in many instances, this waveform is, in turn, bounded by valley width. They also found that valley width appears to inﬂuence the magnitude of the fractal dimension itself. In narrow valleys, it is not possible for the planform to become strongly nonlinear. They discovered in some of their rivers an intermediate scale at which the fractal property changed (so the planform exhibited two distinct scaling ranges), and found that this was determined by meander amplitude. In general, three scaling regions might be deﬁned (Nikora et al., 1993). Near the physical lower limit, the channel often is nonfractal, or ‘‘smooth’’, in its behaviour. This region would normally correspond with the pool-rifﬂe sequence. At larger scales, the pattern passes into a fractal scaling region, which ultimately is limited by larger scale landscape constraints on pattern development. Beyond that, behaviour may be self-afﬁne, so long as the river remains under the inﬂuence of similar governing conditions. Empirical scales for delimiting these domains might be w and Wv, the latter denoting valley ﬂoor width, but entirely unconﬁned channels do not exhibit an inﬁnite fractal range.

Multiple scales in rivers

19

Divided channels present a dramatically different morphology than do singlethread ones but questions related to the scaling of channel pattern remain rather similar. Since the basic building blocks of alluvial channel pattern are, in all cases, bars and pools, this is not surprising. Ashmore (2001) emphasized the afﬁnities between single thread and braided channels and identiﬁed mean braid bar length as a fundamental scale analogous to pool-rifﬂe length (or meander length) in singlethread channels. But a range of bar scales is obviously present. Self-afﬁnity of bar and channel geometry in braided channels is summarized in Paola and FoufoulaGeorgiou (2001). A self-similar range has been identiﬁed by Walsh and Hicks (2002) with upper limit of order Wb, the braidplain width. At the lower end, meaningful channel division would be limited by some multiple of grain size (which would limit the occurrence of aggregate sediment deposits) and by stream competence, hence possibly by relative roughness. At the upper end, a limit in downstream afﬁnity is apt to be set, as for meanders, by topographic space. An interesting study by Sapozhnikov and Foufoula-Georgiou (1997) demonstrates that time scaling across changes in length scale in a developing braided channel network is characterized by tr ¼ l 1=2 r , in which tr and lr are the time and length ratios of the compared examples, respectively. That is, changes at different scales in a self-similar or self-afﬁne channel obey Froude scaling. Within limits posed by the governing conditions – particularly, in this case, limits posed by the boundary materials – comparisons across scales within the channel network constitute classical Froude models of each other. This observation goes some considerable way to explain the strongly suggestive outcomes of many early river modeling exercises that were not formally scaled at all. At the still larger scales of the drainage basin, stream channels are organized into tree-like, hierarchical drainage networks. Since water ﬂows follow the line of steepest accessible descent, the drainage network is subject to topographical constraints set by landscape geometry, bedrock structure, hydroclimate, and drainage basin size. These conditions are mainly externally set and contingent in nature. Nonetheless, fractal self-similarity of drainage networks has been demonstrated (Tarboton et al., 1988; Rodrı´ guez-Iturbe and Rinaldo, 1997). Such appearances can be justiﬁed in terms of the necessarily space-ﬁlling behaviour of channel networks in order to drain the land surface, and can be simulated using various rules for drainage extension or network composition. At the scale of processes in river channels, however, one is concerned with individual conﬂuences and with the distance between successive signiﬁcant conﬂuences (the limit length for a homogeneous reach). Over longer reaches the channel network structure is likely to be determined by landscape history – by inheritance – rather than by contemporary processes. The valley networks within which rivers run have long, complex histories that are not usually dominated by the contemporary river. A new set of landscape scales may be constructed to deﬁne valley networks, but these take us beyond the scope of this paper and will not be pursued.

8.

Ecological scales in rivers

A river appears, superﬁcially, to be a relatively hostile environment for most life. The incessant, strongly turbulent ﬂow requires strong rooting to resist, or signiﬁcant

20

Michael Church

expenditure of energy to withstand. Aquatic organisms seek means to minimize the necessary expenditure of energy to maintain position and to conduct their activities. Beyond anatomical features such as streamlined shape, holdfast mechanisms, and low-resistance surfaces, their evolved behaviour patterns lead them to seek out special environments and to adapt their life-cycle activities to certain seasons and conditions of the river so that energy may be conserved. In these activities, aquatic animals, in particular, demonstrate adaptations at scales that correspond with the physical scales we have been studying (Statzner and Higler, 1986; Hart and Finelli, 1999). In this section, we give only a brief introduction to these adaptations and their scale associations. Many bottom-dwelling aquatic insects are dorso-ventrally ﬂattened in order to be able to live ‘‘under the turbulence’’ in the laminar sublayer. In gravel-bed rivers, this would amount to seeking out the sublayer on individual clasts. Filter-feeding insects, whose foraging behaviour entails intercepting drift material passing in the turbulent ﬂow, and certain scrapers have developed strong claws or hooks in order to be able to position themselves on exposed surfaces. Most preferentially inhabit areas of accelerating ﬂow with favourable pressure gradients, where destabilizing turbulent stresses are smaller. Nevertheless, they inhabit only the lowest few millimeters of ﬂow, where velocities remain relatively small. Many others conduct their lives in the lee of the larger clasts; most live primarily under the surface clasts where they are protected both from strong currents and from open-water predators. The body scaling of these insects is similar to the sublayer thickness, ys, or to characteristic bed pore size. The latter is an ecologically important measure that we have heretofore ignored, which is closely allied with D. For closely packed sediments pore opening, D0 E0.4D, but D will be some relatively small fraction in the size distribution. Accordingly, embedded substrate – a gravel bed with abundant ﬁnes blocking the clast interstices – represents poor habitat. More generally, the connectivity of subsurface openings is critical both for incubation of buried eggs and for the life patterns of hyporheic fauna. Like surface networks, there is accumulating evidence for the fractal nature of pore size openings and connections. Important animal interactions occur at scales comparable with bed grain size, since individual clasts may deﬁne signiﬁcant habitat units for insects, for the fry of some ﬁsh species, and even for some larger ﬁshes over suitably coarse substrates. Competitive interactions and density variations may be evident only at the small scales that represent actual habitat units (Downes et al., 1993). Stable clasts and clast aggregations are particularly important as refugia during strong ﬂows (Biggs et al., 1997; Matthaei et al., 2000). At a larger scale, then, stable rifﬂes are disproportionately important sites. Seasonal and/or life-cycle movements of many invertebrates are controlled at the turbulent macroscale or depth scale of the river, or at the scale of pool-rifﬂe spacing. Many species, particularly crawlers, move only on the order of meters over periods of weeks (Malmqvist, 2002). Life-cycle requirements lead some species to move from rifﬂe to pool habitats, from rifﬂes to shallows, or to off-channel areas (Bilton et al., 2001). For others, seasonal movements represent a survival strategy in face of strong seasonal variations in ﬂow. Some species follow a shifting zone of constant depth, while others seek relatively constant velocity or shear stress (Rempel et al., 1999).

Multiple scales in rivers

21

Locomotive abilities inﬂuence the movements of many species but, of course, some species travel long distances by drifting in the turbulent ﬂow. Fishes are altogether more mobile than most waterborne insects. Habitat preferences have been categorized directly with depth and ﬂow velocity (Stalnaker et al., 1989), but these preferences are usually expressed by the ﬁsh through selection for speciﬁc morphological units of the channel. Such units may be associated with grainto rifﬂe-scaled features in gravel-bed rivers and are often associated with irregularities of the stream edge which afford suitable water, hiding room, and foraging possibilities. A wide range of such local habitats has been identiﬁed (Bisson et al., 1981; Padmore et al., 1998; Church et al., 2000) with typical dimensions comparable with w. In small channels, they may largely be equivalent to channel morphological units, but in large channels, they are much smaller units. When they travel, ﬁsh interact signiﬁcantly with ﬂow at the turbulent macroscale. It has recently been demonstrated (Liao et al., 2003) that upstream-swimming ﬁsh use vertically oriented eddies – typically ‘‘vortex streets’’ – to reduce swimming energy requirements by moving always into the upstream-directed side of the vortices. In effect, they slalom between successive eddies. Because vortex streets are shed from bluff bodies and bank protrusions, ﬁsh often swim in narrowly deﬁned ‘‘pathways’’ along steep channel edges or from object to object. Long distance swimmers take advantage of larger eddy structures as well. Upstream migrating salmonids systematically cross the stream to swim in the ‘‘inner’’ bank slack water or separation eddy. Aquatic communities are structured at the habitat scale and, more loosely, at the river reach scale. Some species move more readily than others, according to foraging strategy, but some individuals of a particular species may also be more mobile than others. Channel pattern types are apt to be ecologically distinctive because they offer distinctive physical conditions and ranges of variance (e.g., Davey and Lapointe, 2007). Communities are deﬁned, then, within the reach scales of the channels that hold them. Aquatic communities survive, however, within larger physical systems that are strongly directed by water and nutrient ﬂow, and so food webs may extend to span an entire catchment (Woodward and Hildrew, 2002). Within the catchment, there are strong contrasts in nutrient recruitment, with abundant addition and possibly initial processing of carbonaceous materials in headwaters and successive stages of instream processing and reproduction of nutrient materials downstream. Gravel-bed channels that are found in the mid-reaches of many upland draining catchments are often richest in aquatic resources as the consequence of abundant nutrient delivery, the persistence of a varied range of habitats created by the diversity of gravel accumulations, and the moderate but persistent level of disturbance driven by low rates of bed material transport.

9.

Discussion

In this paper, I have sought to identify signiﬁcant scales in length and time of processes that lend character to gravel-bed rivers. Attention has been focused on classical scales associated with the ﬂow at turbulent and channel scales, and with channel morphology at unit and reach scales. Intrinsic turbulent length scales include

22

Michael Church

a well-established microscale, n/u, and a macroscale, uNTEou4T, in which T is the timescale of large coherent structures. But the latter remains controversial, in part because the true nature of large coherent structures in the turbulent shear ﬂow remains uncertain. The smallest scale of morphological signiﬁcance in gravel-bed rivers is the boundary grain scale, D, which is seen to scale both resistance to ﬂow and processes associated with bed material transport. Grain sizes fall within the range of scales spanned by turbulence, and so the bed elements interact directly with the ﬂow to create turbulent structures. For the channel, evident length scales appear to include d, w, and xp, the latter being understood to be a characteristic bed particle step length typically approaching l/2. These are all simple geometric scales and, since boundary materials constrain channel shape, only xp might be intrinsic. It appears to be related to dynamical features of the ﬂow, but the nature of those features remains unclear. A special case occurs when D-d (D/d-1.0) since, then, individual grains become signiﬁcant morphological elements of the channel. In this case, the dominant morphological scale of the channel is set by the relation between grain size and channel gradient. Most channel reaches are constrained by the valley within which the river runs. Some rivers escape notable lateral constraints on alluvial fans and when ﬂowing into large sedimentary basins, but there is usually a length constraint in such cases. Accordingly, the scales of repeating patterns in river channels are limited by w and by Wv, within the limits of which there may be no singular morphological scale. A strongly impressed scale, however, is meander wavelength or, equivalently, dominant braid bar wavelength. Some such scale is present even in fully conﬁned channels. These scales are closely set by xp, and so do not convey additional information (but they are much more easily measured). A diagram that associates dominant features of the ﬂow and morphology of rivers with characteristic dimensions of length and time is given as Fig. 1.5. There is an increasingly obvious fundamental problem underlying the selection of all of the morphologically signiﬁcant scales. Each scale is more and more obviously not a singular reference dimension, but is merely representative in some way of a range of phenomena. We observe the scaling range of turbulent structures (Fig. 1.1), limited at the lower end by the near-boundary generating mechanism and, seemingly, at the upper end by the size of the ‘‘container’’. Hence, the association of large coherent structures with ﬂow depth. Much of the intervening scaling range may be occupied by shearline eddies created within the ﬂow. However the scaling range is created, we are left with scales that delimit the range of turbulent ﬂow structures, while the interesting phenomena associated with energy transfer and dissipation, and with sediment entrainment and transport, occur at all scales within that range. Bed material also presents a range of sizes – a range created by cycles of entrainment and deposition of the sediments. This sorting process does not possess strict limits. Entrainment and disentrainment can be thought of as distinctly leaky ﬁlters operating on the streambed, leaky in part, at least, because the bed itself interferes with the transport over it so that grains are variously hidden from entrainment, or trapped and prematurely disentrained. Grain resistance to ﬂow, which helps to set mean ﬂow velocity, is disproportionately affected by the large sizes present on the bed, whereas sizes apt to be transported (hence suitable to scale the transport

Multiple scales in rivers

23 -1

0 10

1012

m

a

-1

drainage

tectonics

0

1010

neotectonics

m

y da

10

basin -1

=ν

108

reach

0

T

-1

1 year

L

2

morphological scale ranges

TIME SCALE (m)

106

molecular diffusion 1 day

104

channel subchannel morpholog y

1 hour

10

channel secondary

102

m

s

flows

1 minute

turbulent

dynamic scale ranges LT

-2 =

g

100

10-2

10-2

100

102 104 LENGTH SCALE (m)

106

Figure 1.5. Representation of scale ranges in L– T space. Range limits are set by the scales of molecular diffusion and gravity waves. Trajectories in this space deﬁne characteristic system velocities. A typical velocity of 1 m s1 for water intersects turbulent and channel scales, while a long-term virtual velocity of 102 m a1, perhaps typical of mobile sediment, intersects scales from channel to tectonics. Subchannel morphology (granular accumulations such grain clusters, or ripples and dunes) would be associated with velocities, measured in the short-term, on the order of 10 m day1.

process) are, in gravels, usually much smaller. Grains on the streambed can form aggregate structures and do form larger, bar-form accumulations that add additional resistance to ﬂow, so there is not a true simple scale here at all. A roughness length that incorporates the total resistance to ﬂow can be back-calculated from knowledge of stream velocity, but it is a conceptual length scale only, not susceptible of independent measurement.

24

Michael Church

Two scales that apparently are clearly deﬁned are channel depth and channel width. They are set by the governing conditions of water supply, boundary materials, and stream gradient. But complications lurk here, as well, because ﬂows in rivers are more or less continuously varied, so the problem arises of deﬁning an appropriate reference ﬂow, and with it a ‘‘regime’’ width and depth. This is a very old problem, but it is no closer to resolution for that. Experimental evidence suggests that grain paths in transport do have a characteristic length scale, and the reason for it must be associated with some aspect of the ﬂow, as is seen by the occurrence of equivalent length scales impressed into ice or bedrock by purely erosional ﬂows. The association appears to lie with large, streamwise vortical structures that are themselves shaped by the channel and ultimately pinned by the channel deformation they induce (hence, are not, properly, turbulent structures), but the chain of cause and effect has not entirely been worked out and the observed water circulation in alluvial channels could be a trailing phenomenon. Furthermore, it is evident that the characteristic path length, giving rise to the poolrifﬂe sequence, or to the dominant meander dimension, is again a limit value. Actual path lengths cover a wide distribution and we commonly observe only the integral outcome of the transport process. Further elaboration of this picture will require more careful deﬁnition of timescales and particle step sequences, something that is still difﬁcult because of measurement limitations. Whilst particle path length might be expected to give rise to a well-deﬁned channel morphological scale, it turns out that channel morphological units repeat in selfsimilar (or self-afﬁne) fashion over a range of scales. Why does this happen? In the case of braided channels, which repeat bar and channel features, local variations in sediment transport and sedimentation must be associated with the scale variation. This effect was demonstrated in a simple simulation model by Murray and Paola (1994) and is consistent with measurements by Ashmore (1988) of variable sediment transport at constant ﬂow rate in a braided system. Paola and Foufoula-Georgiou (2001) claim that the effect is the consequence of the nonlinearity of sediment transport relations and connect this behaviour with the characteristic behaviour of self-organized critical systems, which typically store sediment (or energy) up to some critical limit after which they adjust over all possible length scales. There is, of course, an element of contingency introduced at channel scales, within which sediment accumulations persistently reﬂect the recent history of ﬂow and sediment inﬂux, so self-organized conditional stability might be a more apt description of channel state. In single-thread channels, the origin of sediment-driven self-similarity is perhaps more subtle since the physical sediment queue is more nearly linear. Single-thread channels contend with banks of variable strength that may give rise to variations in both channel form and local sediment mobilization, given nonlinear forcing. Certainly, variations in sediment transport have been documented over a range of scales varying from seconds to months or years (summaries are given by Reid et al., 1985, and Gomez, 1991). A consequent range in xp, hence in channel unit lengths scales, would not be surprising in such a circumstance. Beyond that, one must contend with the quite different temporal forcing of water and of sediment supply to stream channels, a circumstance that has been relatively little considered because of the long dominant

Multiple scales in rivers

25

ﬁction that bed material transport in alluvial channels must proceed according to some functionally ﬁxed hydraulic capacity. In the longer term, it appears that recognition of the scaling range of many of the most important features of ﬂow, sedimentation, and morphology of river channels will compel a more detailed examination of almost all aspects of river processes, one consequence of which will be a radical rethinking of what are the fundamental scales, or scale limits, of river processes and forms. I hope that this paper might prompt a more systematic investigation of the issues.

Acknowledgements I thank Paul Jance for producing the drawings on very short notice, Brett Eaton for collaboration on river channel scales, and three reviewers for helpful insights that forced me to clarify many of my statements.

References Aberle, J., Nikora, V., 2006. Statistical properties of armoured gravel bed surfaces. Water Resour. Res. 42, W11414 , doi:10.1029/2005WR004674. Aberle, J., Smart, G.M., 2003. The inﬂuence of roughness structure on ﬂow resistance on steep slopes. J. Hydraul. Res. 41, 259–269. Ackers, P., 1988. Alluvial channel hydraulics. J. Hydrol. 100, 177–204. Ashmore, P.E., 1988. Bedload transport in braided gravel-bed stream models. Earth Surf. Process. Landf. 13, 677–695. Ashmore, P.E., 2001. Braiding phenomena: Statics and kinetics. In: Mosley, M.P. (Ed.), Gravel-Bed Rivers V. New Zealand Hydrological Society, Wellington, New Zealand, pp. 95–114. Beauvais, A.A., Montgomery, D.R., 1996. Inﬂuence of valley type on the scaling properties of river planforms. Water Resour. Res. 32, 1441–1448. Best, J.L., 1993. On the interaction between turbulent ﬂow structure, sediment transport and bedform development: Some considerations from recent experimental research. In: Clifford, N.J., French, J.R., and Hardisty, J. (Eds), Turbulence: Perspectives on Flow and Sediment Transport. Wiley, New York, pp. 61–92. Biggs, B., Duncan, M., Francoeur, S., Meyer, W., 1997. Physical characteristics of microform bed cluster refugia in 12 headwater streams, New Zealand. New Zeal. J. Mar. Freshwat. Research 31, 413–422. Bilton, D.T., Freeland, J.R., Okamura, B., 2001. Dispersal in freshwater invertebrates: Mechanisms and consequences. Annu. Rev. Ecol. Systemat. 32, 159–181. Bisson, P.A., Neilsen, J.L., Palmason, R.A., Grove, L.E., 1981. A system of naming habitat types in small streams, with examples of habitat utilization by slamonids during low stream ﬂow. In: Armentrout, N.B. (Ed.), Acquisition and Utilization of Aquatic Habitat Inventory Information. American Fisheries Society, Western Division, Portland, OR, pp. 291–298. Brayshaw, A.C., 1984. Characteristics and origin of cluster bedforms in coarse-grained alluvial channels. In: Koster, E.J. and Steel, R.J. (Eds), Sedimentology of gravels and conglomerates. Can. Soc. Petr. Geol. Mem, Canada, Vol. 10, pp. 77–85. Bufﬁn-Be´langer, T., Roy, A.G., Kirkbride, A.D., 2000. On large-scale ﬂow structures in a gravel bed river. Geomorphology 32, 417–435. Cantwell, B.J., 1981. Organized motion in turbulent ﬂow. Annu. Rev. Fluid Mech. 13, 457–515. Cellino, M., Graf, W.H., 1999. Sediment-laden ﬂow in open channels under noncapacity and capacity conditions. J. Hydraul. Eng. 125, 455–462.

26

Michael Church

Church, M., Hassan, M.A., 1992. Size and distance of travel of unconstrained clasts on a streambed. Water Resour. Res. 28, 299–303. Church, M., Hassan, M.A., Wolcott, J.F., 1998. Stabilizing self-organized structures in gravel-bed streams: Field and experimental observations. Water Resour. Res. 34, 3169–3179. Church, M., Rempel, L.L., and Rice, S., 2000. Morphological and habitat classiﬁcation of the lower Fraser River gravel-bed reach. Report for the Fraser Basin Council, 77pp. Available at www.geog.ubc.ca/ fraserriver/reports.html Clifford, N.J., 1993. Formation of rifﬂe-pool sequences: Field evidence for an autogenetic process. Sediment. Geol. 85, 39–51. Clifford, N.J., Robert, A., Richards, K.S., 1992. Estimation of ﬂow resistance in gravel-bedded rivers: A physical explanation of the multiplier of roughness length. Earth Surf. Process. Landf. 17, 111–126. Corino, S.R., Brodkey, R.S., 1969. A visual investigation of the wall region in turbulent ﬂow. J. Fluid Mech. 37, 1–30. Davey, C., Lapointe, M., 2007. Sedimentary links and the spatial organization of Atlantic salmon (Salmo salar) spawning habitat in a Canadian Shield river. Geomorphology 83, 82–96. Dement’ev, V.V., 1962. Investigation of pulsations of velocities of ﬂow of mountain streams and of its effect on the accuracy of discharge measurements. Translated (into English) in Sov. Hydrol., No. 6, 588–622, by D.B. Krimgold. Downes, B.J., Lake, P.S., Schreiber, E.S.G., 1993. Spatial variation in the distribution of stream invertebrates: Implications of patchiness for models of community organization. Freshwat. Biol. 30, 119–132. Eaton, B.C., Church, M., 2004. A graded stream response relation for bedload dominated streams. J. Geophys. Res.: Earth Surface 109 (F03011), doi:10.1029/2003JF000062. Eaton, B.C., Church, M., Davies, T.R.H., 2006. A conceptual model for meander initiation in bedload dominated streams. Earth Surf. Process. Landf. 31, 875–891. Eaton, B.C., Church, M., Millar, R.G., 2004. Rational regime model of alluvial channel morphology and response. Earth Surf. Process. Landf. 29, 511–529. Einstein, H.A., 1937. Bedload transport as a probability problem. Dissertation (in German). Translated in English by W.W. Sayre. In: Shen, H.W. (Ed.), Sedimentation. Fort Collins, CO, WRP. Appendix C: 105pp. Garcia, M., Nino, Y., 1993. Dynamics of sediment bars in straight and meandering channels: Experiments on the resonance phenomenon. J. Hydraul. Res. 31 (6), 739–761. Gomez, B., 1991. Bedload transport. Earth Sci. Rev. 31, 89–132. Grass, A.J., 1971. Structural features of smooth and turbulent ﬂow over smooth and rough boundaries. J. Fluid Mech. 50, 233–255. Grifﬁths, G.A., 2003. Downstream hydraulic geometry and hydraulic similitude. Water Resour. Res. 39 (4), doi:10.1029/2002WR001485. Hart, D.D., Finelli, C.M., 1999. Physical-biological coupling in streams: The pervasive effects of ﬂow on benthic organisms. Annu. Rev. Ecol. Systemat. 30, 363–395. Huang, H.Q., Nanson, G.C., 2000. Hydraulic geometry and maximum ﬂow efﬁciency as products of the principle of least action. Earth Surf. Process. Landf. 25, 1–16. Ikeda, S., 1984. Prediction of alternate bar wavelength and height. J. Hydraul. Eng. 110 (14), 371–386. Jackson, R.G., 1976. Sedimentological and ﬂuid-dynamic implications of the turbulent bursting phenomenon in geophysical ﬂows. J. Fluid Mech. 77, 531–560. Keulegan, G.H., 1938. Laws of turbulent ﬂow in open channels. United States National Bureau of Standards. J. Res. 21, 707–741. Kirkbride, A.D., 1993. Observation of the inﬂuence of bed roughness on turbulent structure in depth limited ﬂows over gravel beds. In: Clifford, N.J., French, J.R., and Hardisty, J. (Eds), Turbulence: Perspectives on Flow and Sediment Transport. Wiley, New York, pp. 185–196. Klebanoff, P.S., 1954. Characteristics of turbulence in a boundary layer with zero pressure gradient. NACA Technical Note, 3178. Kline, S.J., Reynolds, W.C., Schraub, F., Runstadtler, P.W., 1967. The structure of turbulent boundary layers. J. Fluid Mech. 30, 741–773.

Multiple scales in rivers

27

Komori, S., Ueda, H., Ogino, F., Mizushina, T., 1982. Turbulence structure and transport mechanism at the free surface in an open channel ﬂow. Int. J. Heat Mass Transfer 25, 513–521. Korchokha, Yu. M., 1968. Investigation of the dune movement of sediments on the Polomet’ River. Translated (into English) in Sov. Hydrol., No. 6, 541–559. Kostaschuk, R.A., Church, M., Luternauer, J.L., 1991. Acoustic images of turbulent ﬂow structures in Fraser River estuary, British Columbia. In: Current Research Part E. Geol. Surv. Can. Paper No. 91-1E: 83–90. Lapointe, M., 1992. Burst-like sediment suspension events in a sand bed river. Earth Surf. Proc. Landf. 17, 253–270. Laronne, J.B., Carson, M.A., 1976. Interrelationships between bed morphology and bed-material transport for a small, gravel-bed channel. Sedimentology 23, 67–85. Leliavsky, S., 1956. An introduction to ﬂuvial hydraulics. Constable, London, 257pp. Leopold, L.B., Maddock, T. Jr., 1953. The hydraulic geometry of stream channels and some physiographic implications. US Geol. Surv. Prof. Pap. 252, 57. Leopold, L.B., Wolman, M.G., 1960. River meanders. Geol. Soc. Am. Bull. 71, 769–794. Leopold, L.B., Wolman, M.G., Miller, J.P., 1964. Fluvial processes in geomorphology. W.H. Freeman, San Francisco, CA, 522pp. Lewin, J., 1976. Initiation of bed forms and meanders in coarse-grained sediment. Geol. Soc. Am. Bull. 87, 281–285. Liao, J.C., Beal, D.N., Lauder, G.V., Triantafyllou, M.S., 2003. Fish exploiting vortices decrease muscle activity. Science 302, 1566–1569. Makaveev, V.M., 1964. Interaction of turbulent ﬂow with the underlying surface and its formation. Translated (into English) in Sov. Hydrol., No. 3, 224–240 by D.B.Krimgold. Malmqvist, B., 2002. Aquatic invertebrates in riverine landscapes. Freshwat. Biol. 47, 679–694. Mandelbrot, B.B., 1977. Fractals: Form, chance and dimension. W.H. Freeman, New York, 365pp. Matthaei, C., Arbuckle, C., Townsend, C., 2000. Stable stones as refugia for invertebrates during disturbance in a New Zealand stream. J. N. Am. Benthol. Soc. 19, 82–93. McQuivey, R.S., 1973. Summary of turbulence data from rivers, conveyance channels, and laboratory ﬂumes. US Geol. Surv., Prof. Pap. 802-B, 66pp. Murray, A.B., Paola, C., 1994. A cellular model of braided rivers. Nature 371, 54–57. Nikora, V.I., 1991. Fractal structures of river planforms. Water Resour. Res. 27, 1327–1333. Nikora, V.I., Goring, D.G., Biggs, B.F., 1998. On gravel-bed roughness characterization. Water Resour. Res. 34, 517–527. Nikora, V.I., Habersack, H., Huber, T., McEwan, I., 2002. On bed particle diffusion in gravel bed ﬂows under weak bed load transport. Water Resour. Res. 38 (6), doi:10.1029/2001WR000513. Nikora, V.I., Sapozhnikov, V.B., Noever, D.A., 1993. Fractal geometry of individual river channels and its computer simulation. Water Resour. Res. 29, 3561–3568. Nowell, A.R.M., Church, M., 1979. Turbulent ﬂow in a depth-limited boundary layer. J. Geophys. Res. 84, 4816–4824. Padmore, C.L., Newson, M.D., Charlton, M.E., 1998. Instream habitat in gravel-bed rivers: Identiﬁcation and characterization of biotopes. In: Klingeman, P.C., Beschta, R.L., Komar, P.D., and Bradley, J.B. (Eds), Gravel-Bed Rivers in the Environment. Water Resources Publications, Highlands Ranch, CO, pp. 345–364. Paiement-Paradis, G., Bufﬁn-Be´langer, T., Roy, A.G., 2003. Scalings for large turbulent ﬂow structures in gravel-bed rivers. Geophys. Res. Lett. 30 (14), doi:10.1029/2003GL017553. Paola, C., Foufoula-Georgiou, E., 2001. Statistical geometry and dynamics of braided rivers. In: Mosley, M.P. (Ed.), Gravel-Bed Rivers V. New Zealand Hydrological Society, Wellington, New Zealand, pp. 47–71. Parker G. and Anderson A.G., 1975. Modelling of meandering and braiding in rivers. Proceedings of the ASCE Modelling Symposium, American Society of Civil Engineers, San Francisco, CA, pp. 575–591. Pyrce, R.S., Ashmore, P.E., 2003a. The relation between particle path length distributions and channel morphology in gravel-bed streams: A synthesis. Geomorphology 56, 167–187.

28

Michael Church

Pyrce, R.S., Ashmore, P.E., 2003b. Particle path length distributions in meandering gravel bed streams: Results from physical models. Earth Surf. Process. Landf. 28, 951–966. Rao, K.N., Narasimha, R., Badri Narayanan, A.B., 1971. Bursting in a turbulent boundary layer. J. Fluid Mech. 48, 401–430. Reid, I., Frostick, L.E., Layman, J.T., 1985. The incidence and nature of bedload transport during ﬂood ﬂows in coarse-grained alluvial channels. Earth Surf. Process. Landf. 10, 33–44. Rempel, L.L., Richardson, J.S., Healey, M.C., 1999. Flow refugia for benthic macroinvertebrates during ﬂooding of a large river. J. N. Am. Benthol. Soc. 18, 34–48. Rodrı´ guez-Iturbe, I., Rinaldo, A., 1997. Fractal river basins. Cambridge University Press, Cambridge, pp. 547. Rouse, H., 1965. Critical analysis of open channel resistance. American Society of Civil Engineers. Proc., J. Hydraul. Div. 91 (HY4), 1–25. Roy, A.G., Bufﬁn-Be´langer, T., Lamarre, H., Kirkbride, A.D., 2004. Size, shape and dynamics of largescale turbulent ﬂow structures in a gravel-bed river. J. Fluid Mech. 500, 1–27. Sapozhnikov, V.B., Foufoula-Georgiou, E., 1996. Self-afﬁnity in braided rivers. Water Resour. Res. 32, 1429–1439. Sapozhnikov, V.B., Foufoula-Georgiou, E., 1997. Experimental evidence of dynamic scaling and indications of self-organized criticality in braided rivers. Water Resour. Res. 33, 1983–1991. Shvidchenko, A.B., Pender, G., 2001. Macroturbulent structure of open-channel ﬂow over gravel beds. Water Resour. Res. 37, 709–719. Smart, G.M., Duncan, M.J., Walsh, J.M., 2002. Relatively rough ﬂow resistance equations. J. Hydraul. Eng. 128, 568–578. Snow, R.S., 1989. Fractal sinuosity of stream channels. Pure Appl. Geophys. 131, 99–109. Speight, G.H., 1965. Meander spectra of the Angabunga River. J. Hydrol. 3, 1–15. Stalnaker, C.B., Milhous, R.T., and Bovee, K.D., 1989. Hydrology and hydraulics applied to ﬁshery management in large rivers. In: Dodge, D.P. (Ed.), Proceedings of the International Large Rivers Symposium, Canada Department of Fisheries and Oceans, Ottawa, Canadian Special Publication of Fisheries and Aquatic Sciences, Vol. 106, pp. 13–30. Statzner, B., Higler, B., 1986. Stream hydraulics as a major determinant of benthic invertebrate zonation patterns. Freshwat. Biol. 16, 127–139. Tarboton, D.G., Bras, R.L., Rodrı´ guez-Iturbe, I., 1988. The fractal nature of river networks. Water Resour. Res. 24, 1317–1322. Taylor, G.I., 1935. The statistical theory of turbulence, I–IV. Proc. R. Soc. London A135, 421–511. Thompson, A., 1986. Secondary ﬂows and the pool-rifﬂe unit: A case study of the processes of meander development. Earth Surf. Process. and Landf. 11, 631–641. Townsend, A.A., 1970. Entrainment and the structure of turbulent ﬂow. J. Fluid Mech. 41, 13–46. Townsend, A.A., 1976. The structure of turbulent shear ﬂow. 2nd ed. Cambridge University Press, Cambridge, 429pp. van Rijn, L.C., 1982. Equivalent roughness of alluvial bed. J. Hydraul. Eng. 128, 1215–1218. Velikanov, M.A., 1949. Dynamics of channel ﬂows. Translated in Russian in 1954. Leningrad, Gidrometeoizdat, Trudy Gos. Izd. Teckhn-teor. Lit. 1. Walsh, J., Hicks, D.M., 2002. Braided channels: Self-similar or self-afﬁne? Water Resour. Res. 38(6). doi:10.1029/2001WR000749. Wilcock, P.R., McArdell, B.W., 1993. Surface-based fractional transport rates: Mobilization thresholds and partial transport of a sand-gravel sediment. Water Resour. Res. 29, 1297–1312. Wilcock, P.R., McArdell, B.W., 1997. Partial transport of a sand/gravel sediment. Water Resour. Res. 33, 233–245. Wilcock, P.R., Southard, J.B., 1989. Bed-load transport of mixed size sediment. Water Resour. Res. 25, 1629–1641. Woodward, G., Hildrew, A.G., 2002. Food web structure in riverine landscapes. Freshwat. Biol. 47, 777–798. Yalin, M.S., 1977. Mechanics of Sediment Transport. 2nd edn. Pergamon, Toronto, 298pp. Yalin, M.S., 1992. River mechanics. Pergamon, Tarrytown, NY, 219pp.

Multiple scales in rivers

29

Discussion by C.D. Rennie The author should be commended for presenting a stimulating review of the full spectrum of scales in rivers. Based on the work of Pyrce and Ashmore (2003a, 2003b, 2005), the author suggests that meander wavelength scaling in rivers is fundamentally related to sediment transfer. Pyrce and Ashmore (2003b) observed that particle step length during channel forming ﬂows was symmetrically distributed with a mean equal to the pool-bar spacing. Sediment was eroded from pool locations and deposited on the subsequent bar, suggesting that sediment transfer determines the meander wavelength scale. However, it must be recognized that Pyrce and Ashmore observed particle step lengths after pool-bar meander morphology was established. It is entirely sensible that sediment was eroded from zones of ﬂow convergence in the pool and deposited on the subsequent zone of ﬂow divergence on the bar. This does not necessarily mean that the meander geometry was established by the sediment transfer. Flow in a sinuous channel consists of periodically reversing helical ﬂow characterized by twin surface convergent helical cells of alternating strength. The mechanics of meander initiation has engrossed researchers for over a century (Thomson, 1878). The essential question is if meandering is a result of the ﬂow ﬁeld, sediment transport, or an interaction of the two. A meander initiation theory must explain how localized points of erosional bank attack alternate from side to side of the channel. An argument in favour of ﬂow ﬁeld theories is that meandering is observed in other ﬂuid ﬂows in the absence of sediment transport, such as ocean currents (Leopold and Wolman, 1960), supraglacial meltwater channels (Leopold and Wolman, 1960), water rivulets on inclined smooth plates (Gorycki, 1973), channels in sedimentary deposits of deep sea turbidity currents (Coterill et al., 1998; Parker, 1998), pollutant plumes (Etling, 1990), and tropical cyclone paths (Holland and Lander, 1993). This general tendency to meander suggests that meandering is a fundamental characteristic of ﬂuid ﬂow, associated with shear and vorticity production at density interfaces. Admittedly, however, it is presumptuous to assume that similar meander form in these ﬂows suggests similar underlying processes (Knighton, 1998). Parker (1976) argued that in all meandering ﬂows, a ‘third effect’ beyond potential (inertial and gravitational) and frictional forces is required to cause meandering (e.g., sediment transport, Coriolis acceleration, heat differences, or surface tension). However, it appears that this hypothesis remains largely untested. In general, meander development theories based on the ﬂow ﬁeld (e.g., Einstein and Shen, 1964; Quick, 1974), or on ﬂow–sediment interactions (e.g., the bar-bend models of Johannesson and Parker, 1989 and Seminara and Tubino, 1989), require an initial perturbation in the bed surface to generate reversing helical ﬂow and/or alternating zones of erosion and deposition. In natural rivers, initial perturbations could be generated by large woody debris or large clasts, but spatially differential sediment transport is usually evoked. In the two bar-bend models, feedback mechanisms between the ﬂow and sediment result in the ﬁnal meander form. It seems likely that meandering form in rivers, and thus the meander wavelength scale, is the result of interactions between an initial perturbation, reversing helical ﬂow structure,

30

Michael Church

and sediment transport. Importantly, the experimental results of Pyrce and Ashmore do not necessarily eliminate the role of the ﬂow ﬁeld.

References Coterill, K., Coleman, J., Marotta, D., et al., 1998. Sinuosity in submarine channels; scale and geometries in seismic and outcrop indicating possible mechanisms for deposition (abstract). Am. Ass. Pet. Geol. Bull. 82 (10), 1904. Einstein, H.A., Shen, H.W., 1964. A study on meandering in straight alluvial channels. J. Geophys. Res. 69 (24), 5239–5247. Etling, D., 1990. On plume meandering under stable stratiﬁcation. Atmos. Environ. 24A (8 Part 2), 1979–1985. Gorycki, M.A., 1973. Hydraulic drag: A meander initiating mechanism. Bull. Geol. Soc. Am. 84, 175–186. Holland, G.J., Lander, M., 1993. The meandering nature of tropical cyclone tracks. J. Atmos. Sci. 50 (9), 1254–1266. Johannesson, H., Parker, G., 1989. Linear theory of river meanders. In: Ikeda, S. and Parker, G. (Eds), River Meandering. American Geophysical Union, Washington, DC, pp. 181–213. Knighton, D., 1998. Fluvial forms and processes: A new perspective. Arnold, London. Leopold, L.B., Wolman, M.G., 1960. River meanders. Geol. Soc. Am. Bull. 71 (6), 769–793. Parker, G., 1976. On the cause and characteristic scales of meandering and braiding in rivers. J. Fluid Mech. 76, 457–480. Parker, G., 1998. Flow and deposits of turbidity currents and submarine debris ﬂows (abstract). Am. Ass. Pet. Geol. Bull. 82 (10), 1948–1949. Pyrce, R.S., Ashmore, P.E., 2003a. Particle path length distributions in meandering gravel-bed streams: Results from physical models. Earth Surf. Process. Landf. 28 (9), 951–966. Pyrce, R.S., Ashmore, P.E., 2003b. The relation between particle path length distributions and channel morphology in gravel-bed streams: A synthesis. Geomorphology 56 (1–2), 167–187. Pyrce, R.S., Ashmore, P.E., 2005. Bedload path length and point bar development in gravel-bed river models. Sedimentology 52 (4), 839–857. Quick, M.C., 1974. Mechanism for streamﬂow meandering. J. Hydraul. Eng. 100 (HY6), 741–753. Seminara, G., Tubino, M., 1989. Alternate bars and meandering: Free, forced and mixed interactions. In: Ikeda, S. and Parker, G. (Eds), River Meandering. American Geophysical Union, Washington, DC, pp. 267–320. Thomson, J., 1878. On the origin of windings of rivers in alluvial plains, with remarks on the ﬂow of water round bends in pipes. Proc. R. Soc. London, Ser. A. 25, 114–127.

Discussion by A. Roy I would like to raise three points concerning the turbulent scale. The author questions the usefulness of the integral length scale (LE) for determining the upper limit of turbulence. Because it is based on the integration of the autocorrelation function of the whole velocity signal, LE represents the scale of both the turbulent events and the ambient ﬂuid. LE provides a very conservative estimate of the size of the turbulent ﬂow structures. Our data from a range of gravel-bed rivers and of ﬂow conditions show that LE for the streamwise velocity component averages around 0.9d with a standard deviation of 0.3. This scaling is the lowest estimate of eddy size when compared to values obtained from other methods that emphasize the scale of individual events. These latter methods should be preferred in establishing the maximum scale of turbulent eddies.

30

Michael Church

and sediment transport. Importantly, the experimental results of Pyrce and Ashmore do not necessarily eliminate the role of the ﬂow ﬁeld.

References Coterill, K., Coleman, J., Marotta, D., et al., 1998. Sinuosity in submarine channels; scale and geometries in seismic and outcrop indicating possible mechanisms for deposition (abstract). Am. Ass. Pet. Geol. Bull. 82 (10), 1904. Einstein, H.A., Shen, H.W., 1964. A study on meandering in straight alluvial channels. J. Geophys. Res. 69 (24), 5239–5247. Etling, D., 1990. On plume meandering under stable stratiﬁcation. Atmos. Environ. 24A (8 Part 2), 1979–1985. Gorycki, M.A., 1973. Hydraulic drag: A meander initiating mechanism. Bull. Geol. Soc. Am. 84, 175–186. Holland, G.J., Lander, M., 1993. The meandering nature of tropical cyclone tracks. J. Atmos. Sci. 50 (9), 1254–1266. Johannesson, H., Parker, G., 1989. Linear theory of river meanders. In: Ikeda, S. and Parker, G. (Eds), River Meandering. American Geophysical Union, Washington, DC, pp. 181–213. Knighton, D., 1998. Fluvial forms and processes: A new perspective. Arnold, London. Leopold, L.B., Wolman, M.G., 1960. River meanders. Geol. Soc. Am. Bull. 71 (6), 769–793. Parker, G., 1976. On the cause and characteristic scales of meandering and braiding in rivers. J. Fluid Mech. 76, 457–480. Parker, G., 1998. Flow and deposits of turbidity currents and submarine debris ﬂows (abstract). Am. Ass. Pet. Geol. Bull. 82 (10), 1948–1949. Pyrce, R.S., Ashmore, P.E., 2003a. Particle path length distributions in meandering gravel-bed streams: Results from physical models. Earth Surf. Process. Landf. 28 (9), 951–966. Pyrce, R.S., Ashmore, P.E., 2003b. The relation between particle path length distributions and channel morphology in gravel-bed streams: A synthesis. Geomorphology 56 (1–2), 167–187. Pyrce, R.S., Ashmore, P.E., 2005. Bedload path length and point bar development in gravel-bed river models. Sedimentology 52 (4), 839–857. Quick, M.C., 1974. Mechanism for streamﬂow meandering. J. Hydraul. Eng. 100 (HY6), 741–753. Seminara, G., Tubino, M., 1989. Alternate bars and meandering: Free, forced and mixed interactions. In: Ikeda, S. and Parker, G. (Eds), River Meandering. American Geophysical Union, Washington, DC, pp. 267–320. Thomson, J., 1878. On the origin of windings of rivers in alluvial plains, with remarks on the ﬂow of water round bends in pipes. Proc. R. Soc. London, Ser. A. 25, 114–127.

Discussion by A. Roy I would like to raise three points concerning the turbulent scale. The author questions the usefulness of the integral length scale (LE) for determining the upper limit of turbulence. Because it is based on the integration of the autocorrelation function of the whole velocity signal, LE represents the scale of both the turbulent events and the ambient ﬂuid. LE provides a very conservative estimate of the size of the turbulent ﬂow structures. Our data from a range of gravel-bed rivers and of ﬂow conditions show that LE for the streamwise velocity component averages around 0.9d with a standard deviation of 0.3. This scaling is the lowest estimate of eddy size when compared to values obtained from other methods that emphasize the scale of individual events. These latter methods should be preferred in establishing the maximum scale of turbulent eddies.

Multiple scales in rivers

31

As noted by the author, deﬁning unambiguously a true turbulent event (e.g., ejection) from velocity time series is a prerequisite to any scaling or frequency analysis. It is interesting to note, however, that the detection of turbulent events from velocity records may be quite robust. It is known that the application of various methods used to detect individual ﬂow structures to velocity data will yield different results. In spite of this variability and using thresholds for each method that have been developed from studies in laboratory ﬂumes, it appears that the frequency of events in gravel-bed rivers remains relatively constant among the various detection schemes. For instance, Roy et al. (1996) reported that average bursting frequency (T) for four commonly used burst detection schemes varies between 0.30 and 0.35 event per second. In gravel-bed rivers, bursting frequency is not very sensitive to height above the bed. This suggests that the phenomenon under study may be quite robust in its global properties or that the large-scale events that dominate the turbulent ﬂow ﬁeld in gravel-bed rivers are equally well detected by the schemes. It is important, however, that detection schemes be used consistently and that similar threshold values be applied across various studies. The selection of adequate thresholds may be guided by ﬂow visualization. Bursting frequency in gravel-bed rivers is often quite low with reported values typically between 0.2 and 0.6 depending on the thresholds used in the detection schemes. This indicates that the turbulent structures are quite large. Using Rao et al.’s formula, such low T values would imply that measurements were taken in shallow and/or fast currents. If one supposes that the size of the ejections scale with depth (say 5d) and one samples ﬂows of different depth but with similar ﬂow velocity, then T as estimated by uNT ¼ 5d would increase with increasing depth while the size of the structures would also increase. If larger structures take more time than smaller ones to fully develop and if they advect roughly at the mean ﬂow velocity, it is difﬁcult to imagine that bursting frequency would also increase. It seems to me that it may be difﬁcult to reconcile both scalings as in L ¼ 5d ¼ uNT in many contexts encountered in gravel-bed rivers.

References Roy, A.G., Bufﬁn-Be´langer, T., Deland, S., 1996. Scales of turbulent coherent ﬂow structures in gravel-bed rivers. In: Ashworth, P.J., Bennett, S.J., Best, J.L., and McLelland, S.J. (Eds), Coherent Flow Structures in Open Channels. Wiley, Chichester, pp. 147–164.

Reply by the author I thank Rennie and Roy for their helpful ampliﬁcation of some points in the paper. Together, the discussions emphasize that the essential connections between the ﬂow ﬁeld and the resulting channel morphology remain an important unresolved problem. Rennie points out that the demonstration by Pyrce and Ashmore (2003b) of a preferred rifﬂe–rifﬂe step length for sediment grain movements was made with

Multiple scales in rivers

31

As noted by the author, deﬁning unambiguously a true turbulent event (e.g., ejection) from velocity time series is a prerequisite to any scaling or frequency analysis. It is interesting to note, however, that the detection of turbulent events from velocity records may be quite robust. It is known that the application of various methods used to detect individual ﬂow structures to velocity data will yield different results. In spite of this variability and using thresholds for each method that have been developed from studies in laboratory ﬂumes, it appears that the frequency of events in gravel-bed rivers remains relatively constant among the various detection schemes. For instance, Roy et al. (1996) reported that average bursting frequency (T) for four commonly used burst detection schemes varies between 0.30 and 0.35 event per second. In gravel-bed rivers, bursting frequency is not very sensitive to height above the bed. This suggests that the phenomenon under study may be quite robust in its global properties or that the large-scale events that dominate the turbulent ﬂow ﬁeld in gravel-bed rivers are equally well detected by the schemes. It is important, however, that detection schemes be used consistently and that similar threshold values be applied across various studies. The selection of adequate thresholds may be guided by ﬂow visualization. Bursting frequency in gravel-bed rivers is often quite low with reported values typically between 0.2 and 0.6 depending on the thresholds used in the detection schemes. This indicates that the turbulent structures are quite large. Using Rao et al.’s formula, such low T values would imply that measurements were taken in shallow and/or fast currents. If one supposes that the size of the ejections scale with depth (say 5d) and one samples ﬂows of different depth but with similar ﬂow velocity, then T as estimated by uNT ¼ 5d would increase with increasing depth while the size of the structures would also increase. If larger structures take more time than smaller ones to fully develop and if they advect roughly at the mean ﬂow velocity, it is difﬁcult to imagine that bursting frequency would also increase. It seems to me that it may be difﬁcult to reconcile both scalings as in L ¼ 5d ¼ uNT in many contexts encountered in gravel-bed rivers.

References Roy, A.G., Bufﬁn-Be´langer, T., Deland, S., 1996. Scales of turbulent coherent ﬂow structures in gravel-bed rivers. In: Ashworth, P.J., Bennett, S.J., Best, J.L., and McLelland, S.J. (Eds), Coherent Flow Structures in Open Channels. Wiley, Chichester, pp. 147–164.

Reply by the author I thank Rennie and Roy for their helpful ampliﬁcation of some points in the paper. Together, the discussions emphasize that the essential connections between the ﬂow ﬁeld and the resulting channel morphology remain an important unresolved problem. Rennie points out that the demonstration by Pyrce and Ashmore (2003b) of a preferred rifﬂe–rifﬂe step length for sediment grain movements was made with

32

Michael Church

developed rifﬂes. It is difﬁcult to see how it could be otherwise since the rifﬂes themselves are simply the consequence of such a preferred scale for grain displacements. The problem recurs in all situations in which more or less regular accumulations are observed in sediment transport systems – for example, in the production of aeolian ripples. Rennie reviews farther the largely speculative literature on the possible origin of channel scale regular deformation. I agree that the origin must lie in the ﬂow for reasons that we both have both adduced. But whilst our mathematical theories require some initial, imposed perturbation, it is not at all clear that nature does. Roy has, I think, misunderstood my discussion of the turbulent integral length scale. I do not intend to criticize its role in describing the syndrome of turbulence, but to question whether it can illuminate the undoubted connection between the ﬂow and the consequent scales of ﬂuvial morphology. In any case, as the balance of his discussion makes clear, the signiﬁcant scaling of turbulent events itself remains a subject for additional work.

Gravel-Bed Rivers VI: From Process Understanding to River Restoration H. Habersack, H. Pie´gay, M. Rinaldi, Editors r 2008 Elsevier B.V. All rights reserved.

33

2 Gravel-bed rivers at the reach scale Rob Ferguson

Abstract Reaches extending for tens or hundreds of channel widths along a river are normally seen as stable and homogeneous from a regime perspective, but other approaches emphasise channel change and within-reach spatial variability in ﬂow and physical habitat. Tensions between different perspectives are discussed, with particular emphasis on limitations of traditional regime approaches. Cross-cutting issues include reconciling different scales of self-organisation in gravel-bed rivers, the need to treat bed characteristics as a degree of freedom, and within-reach spatial variability and temporal ﬂuctuation. New ways to tackle these and other issues have been enabled by ever-increasing computing power. Non-uniform and/or unsteady ﬂuvial processes can be modelled numerically, and remote-sensing methods have been developed to acquire dense spatially distributed measurements. But neither models nor observations are infallible, and models of different complexity need to be compared and assessed carefully.

1.

Introduction

The term ‘reach’ is widely used in the ﬂuvial literature but seldom deﬁned. It refers to a stretch of river channel not a speciﬁc locality, but a stretch much shorter than the full distance from headwaters to sea. Sometimes the limits of a study are set by data availability or the geographical limits of a management plan, but usually the ‘reach’ is perceived to be homogeneous in some way. My dictionary gives ‘a straight uniform stretch of river’ but this is too restrictive. A possible working deﬁnition of ‘reach scale’ is the length scale over which relevant characteristics of a river remain essentially the same. Thus a reach might have a certain channel pattern, support a certain ecosystem, be incising, or whatever. This implies lower and upper limits to the length of a reach. An individual pool or meander bend is not a reach, because its physical and ecological properties differ systematically from the adjacent rifﬂe or crossover. The lower limit of ‘reach scale’ is therefore the wavelength of bed macroforms: 101 E-mail address: [email protected] ISSN: 0928-2025

DOI: 10.1016/S0928-2025(07)11112-3

34

R. Ferguson

channel widths. The upper limit is the distance over which changes in channel properties become unacceptably large for the purpose of the investigation, for example because major tributaries add signiﬁcant discharge and load. This upper limit depends on both study purpose and network-scale properties of the river system, but is usually 4102 channel widths. Reaches therefore extend over tens or hundreds of widths. This is also the length scale at which disequilibrium due to environmental change or human activity becomes most obvious: channel widening or incision at a single cross section might relate only to bar or bend growth, but systematic change in width or bed elevation over a series of sections tells us the reach is not in equilibrium. More generally, the reach scale is where generic understanding of ﬂuvial processes has to be reconciled with the contingency of particular landscapes and their forcing by climate and human activity. The literature on gravel-bed rivers at reach scale is large, ever-growing, multidisciplinary, and very varied, but four longstanding perspectives can be distinguished:

Engineers, and some geomorphologists, are interested in channel regime: what a channel must be like to remain in equilibrium. Geomorphologists, and many land managers and engineers, are aware of channel change: dynamic aspects of river behaviour such as meander migration, incision, and aggradation. Sedimentologists, geomorphologists and engineers have studied the key process of bedload transport and its interactions with bed characteristics and sediment supply. Ecologists, and subsequently geomorphologists, have focused on physical habitat diversity as a key aspect of in-stream ecology and a prerequisite for successful river restoration. Some riverine ecology deals with phenomena at sub-reach scale (e.g., animal movement within the hyporheic zone) but overall habitat characteristics are a reach property.

I begin this overview by highlighting some inconsistencies and tensions between these traditional perspectives. Things that are emphasised in some are largely ignored in others, suggesting that our science is not fully joined up. Because bedload transport and ﬂoodplain ecology are the subjects of detailed reviews elsewhere in this volume I base my discussion around the regime and channel-change themes, with links to the other two. The rest of the paper is inspired by the view that science is ‘the art of the soluble’ (Medawar, 1967). Recent technical and methodological developments, underpinned by the inexorable rise in computer power and the ingenuity of numerate ﬂuvial scientists, have greatly increased our ability to measure, describe, or predict ﬂuvial phenomena. This makes it possible to describe speciﬁc river reaches in far more detail than before, and to evaluate general explanatory hypotheses which were previously untestable. The key advances relevant to reach-scale research are the rapid development of methods for acquiring spatially distributed data, and the extension of modelling from one-dimensional (1-D, i.e., width-averaged) to 2-D and 3-D situations. I summarise these developments and discuss the problem of deciding what faith we can put in numerical models.

Gravel-bed rivers at the reach scale 2.

35

Some limitations of present theory

Certain tensions between the four perspectives identiﬁed in Section 1 are obvious. For example, the shift of management focus away from the engineering stability of rivers towards their ecological health necessitates a concern for spatial and temporal variability, whereas regime theory deals with averages. Other less-apparent tensions can be identiﬁed by considering three cross-cutting topics: scales of self-organisation, the bed as a degree of freedom, and within-reach spatial and temporal variability. Sections 2.1–2.3 explore these topics, note inconsistencies between traditional approaches, and draw attention to work which may help resolve some of the tensions.

2.1.

Self-organisation in gravel-bed rivers

Rivers are gutters which drain the landscape and transfer sediment from source areas to alluvial basins, but they are gutters with deformable boundaries. Channel conﬁguration controls ﬂow and sediment transport in the short term but is eventually altered by them. Self-organisation occurs at four scales of which the ﬁrst and last do not concern us here: the development of a drainage network, channel regime, planform development, and depth-scaled bedforms. The regime concept is that river channels adjust towards a dynamic equilibrium in which, averaged over a period of years, the channel is just able to convey the water and sediment supplied from upstream. The key adjustable property is bed elevation, which affects transport capacity through channel slope. Gilbert recognised over a century ago that local aggradation or degradation diffuses up and downstream to provide negative feedback. Channel slope can also be adjusted via sinuosity, and transport capacity also depends on width, so the essence of regime theory is to determine the equilibrium slope and width for given values of water discharge, bedmaterial load, and channel boundary materials. Early work by Lacey, Blench, and others used empirical correlations, as did the geomorphological reinvention of the approach by Leopold and Maddock (1953). Since the late 1970s the search has been for a ‘rational’ regime theory that is physically based (e.g., Parker, 1979), though in many cases closed only by an extremal hypothesis (e.g., Eaton et al., 2004). Regime theory helps us understand how a hierarchical system of fewer but bigger gutters works to remove water and sediment from the landscape, but it has several limitations stemming from simplifying assumptions that are made. The ﬁrst issue is that potentially signiﬁcant controls or responses are omitted. Channel pattern is either ignored, with a disclaimer that attention is restricted to straight channels, or valley slope is taken as independent with sinuosity and channel slope free to adjust. The characteristics of the river bed are represented by a single grain diameter which is assumed ﬁxed and independent, contrary to what many bedload specialists think (Section 2.2). The evident role of bank strength has usually been ignored and a versatile treatment of it has only recently been developed (Millar and Quick, 1998; Eaton et al., 2004). A second issue relates to temporal variability (Section 2.3): regime theory, originating as it did in attempts to design irrigation channels to convey steady ﬂows, collapses the ﬂood magnitude-frequency spectrum into a

36

R. Ferguson

dominant discharge. This ignores the non-linearity of the relation between ﬂood magnitude and geomorphic work done; the effects of extreme ﬂoods are instead regarded as disruptions to regime. Third, channels are assumed to be uniform, which ignores the spatial variability created by small-scale self-organisation. This provides a poor basis for ecological research and calls into question the calculations of bedload transport in regime theories (see Section 2.3). One strand of channel-change research is complementary to regime research: the documentation of systematic channel incision, widening, etc over decades or longer. Such changes are normally regarded as shifts to a new regime following some natural or human-induced change in circumstances, and explicable in terms of discharge and sediment supply using qualitative schemes proposed by Lane (1955) and Schumm (1969). Research of this kind has revealed region-wide trends (e.g., Lie´bault and Pie´gay, 2002; Rinaldi, 2003) and provided valuable input to catchment management plans. But interpretations and predictions are subject to two of the same doubts as regime theory (neglected degrees of freedom, possible effect of major ﬂoods) and predictions carry further limitations: regime theory may mis-predict the eventual equilibrium if degrees of freedom are neglected (Wyzga, 2001), and it does not predict rates of adaptation. Case studies of response to major ﬂoods (e.g., Pitlick, 1993) suggest that different properties have different relaxation times so that complex response is possible. In a period of unprecedented human inﬂuence on river basins, and climate variability at a range of timescales, transient change may be the norm not the exception (Lewin et al., 1988). It may be triggered not only by step changes but by shocks (e.g., ﬂoods or landslides), pulses (e.g., gravel extraction), ramp changes (e.g., deforestation), or quasi-cyclic perturbations (e.g., those in ﬂood magnitude induced by atmospheric teleconnections). To move forward from the limitations of regime and regime-change approaches we need quantitative as well as qualitative models of transient change in rivers. Such models will be numerical, not analytical, for ﬂexibility and to deal with multiple, non-linear, and interacting processes. The sediment routing models mentioned below (Section 2.2) provide a starting point, but ought to allow time-varying discharge and to include a sub-model for width adjustment as discussed by Darby et al. (1998). The other tension between regime theory and alternative approaches to channel behaviour relates to scales of self-organisation. Regime and regime-change approaches are implicitly concerned with self-organisation over fairly extended time and space scales. The mechanisms through which this happens are not treated by regime theory, which seeks only an equilibrium solution, but they evidently comprise erosion and deposition of the bed and banks. Regime theory assumes uniform channel geometry, but channel erosion and deposition normally involve self-organisation at a local scale into characteristic repeating bed forms and patterns. The typical outcome in gravel-bed rivers is a bar-pool-rifﬂe morphology, usually accompanied by local sediment sorting and often by a meandering or braided planform. The development of these repeating forms is studied in another branch of the channel-change perspective. It is a fundamentally different kind of self-organisation from what is envisaged in regime change: much more local, and involving positive feedbacks between the non-uniformity of the morphology, ﬂow ﬁeld, and transport ﬁeld. Positive feedback makes for rapid evolution, in contrast to the diffusive character of

Gravel-bed rivers at the reach scale

37

bed aggradation or degradation. As the amplitude of the repeating forms increases negative feedbacks set in. These too are mainly local, as when meander growth is terminated by chute or neck cutoffs. This is reminiscent of self-organised criticality, as Stolum (1998) and Fonstad and Marcus (2003) have noted in connection with meandering and bank erosion respectively. Both morphology and ﬂow remain intrinsically non-uniform at the within-reach scale. As a result, although the style of morphology is usually persistent and can be seen as part of the regime of a reach, it is at best a statistical equilibrium within which there is more or less frequent change at a local scale. This is the main source of disturbance in the riverine ecosystem, and helps sustain high diversity (Ward et al., 2002; Van der Nat et al., 2003). A challenge for future research is to investigate the coupling between the two scales of self-organisation in river reaches. Insight into the development of repeating forms has been gained mainly from laboratory experiments (e.g., Ashmore, 1991) and mathematical stability analysis of small perturbations of a uniform bed (reviewed by Rhoads and Welford, 1991). Both approaches show that barely perceptible irregularities grow into freely migrating quasi-periodic bedforms; ﬂow deﬂection towards the banks leads to incipient meandering which locks migrating forms into place; then there is either a simpliﬁcation of the pattern within a meandering planform, or ongoing instability if ﬁxed bars are dissected by chutes which generate new free bars and a braided planform. Reach-scale channel pattern thus emerges from small-scale non-uniformity, in the way that is characteristic of systems with interacting nonlinear processes. But from the perspective of many ﬁeld workers the channel pattern and aggradational or degradational tendency of a reach control the migration of bars and rifﬂes and determine which locations experience bank erosion (e.g., McLean and Church, 1999). Is the apparent inconsistency between bottom-up and top-down views of channel behaviour just a matter of scale of interest?

2.2.

The bed as a degree of freedom

As well as a tension between regime-analysis and channel-change schools of thought, there is an inconsistency between regime-analysis and bedload-transport schools concerning the degrees of freedom a river has when adapting to change. Rational regime theory uses a characteristic bed grain size in resistance and transport calculations and treats it as ﬁxed, but it is well known that some kinds of river adjustment are accompanied by changes in bed grain size distribution (GSD hereafter). Degrading channels usually become armoured, for example below dams or in ﬂumes with no sediment feed. As with adjustment of width or slope this reduces the imbalance between bedload transport capacity and sediment supply, but through change in critical shear stress (tc) rather than ﬂow shear stress (t). It is instructive to see how powerful this regulating mechanism can be. A river’s width-averaged bedload capacity can be written using the Meyer–Peter & Muller relation as Qs pwt1:5 c

t tc

1:5 1

(2.1)

R. Ferguson

38

in which t p dS, tc p D, and w, d, S, D denote width, depth, slope, and average bed grain diameter respectively. Adopting the Manning–Strickler resistance law yields 0:6 t Q p S0:7 D 0:9 (2.2) tc w For any given initial value of t/tc it is now possible to calculate the sensitivity of Qs to a speciﬁed change in the imposed discharge (Q) or in any of the adjustable channel properties (w, S, or D). Table 2.1 shows the percentage change in bedload capacity with a 10% change in one controlling factor at a time, for a range of typical transport stages. The sensitivity to each control is higher the closer conditions are to the threshold for motion. Grain size has more inﬂuence than any other factor in nearthreshold conditions, and substantial inﬂuence in all conditions typical of gravel-bed rivers. A change in bed GSD implies size selectivity in the entrainment, transport, and deposition of bed material. This occurs because tc varies somewhat with individual grain size in a mixture even though it depends mainly on overall surface properties: mean grain size, the structural arrangement of coarse clasts (Church et al., 1998), and sand content (Wilcock and Crowe, 2003). As ﬂow increases above the threshold for movement the total bedload ﬂux increases rapidly and its GSD converges on the bedsurface GSD. The bedload GSD may match that of the bed as a whole if ﬁner sizes tend to be hidden beneath a coarse surface layer (Parker and Klingeman, 1982). Surface coarsening and associated structural changes are seen as a regulatory mechanism which reduces transport capacity to the extent necessary to match supply, with static degradational armour as the extreme case (Dietrich et al., 1989; Bufﬁngton and Montgomery, 1999). The bed GSD mediates the effect of supply on channel stability, and allows bedload transport and channel evolution to be predicted (Wilcock, 2001). From this perspective, regime theory neglects one of the potential degrees of freedom in reach self-organisation. Bed adjustment is not restricted to coarsening when supply is reduced. There are documented examples of ﬁning after a big increase in supply (e.g., Pitlick, 1993; Montgomery et al., 1999), and downstream ﬁning along concave long proﬁles can be seen as an adjustment that increases distal transport rates and thus reduces aggradation (Ferguson et al., 1996). More generally, any divergence in the total ﬂux of mixed-size bedload can be accommodated not just by aggradation or degradation but also by change in bed GSD and therefore sizefraction availability and overall tc. The timescale for GSD change will often be Table 2.1. Percentage increase in bedload transport capacity for the indicated transport stage t/tc and the indicated percentage change in discharge (Q), width (w), slope (S) or bed grain diameter (D). See text for basis of calculations. t/tc

Q +10%

w

1.2 1.3 1.4 1.5

57 41 32 28

48 31 23 18

10%

S +10%

D

68 48 38 33

72 46 34 26

10%

Gravel-bed rivers at the reach scale

39

shorter than the timescale for signiﬁcant change in bed elevation (Deigaard and Fredsoe, 1978; Ferguson et al., 1996). It is easier to identify this weakness in traditional regime theory than to suggest how the theory could be extended. The representative grain size cannot become a variable without adding another process constraint, and there is no obvious candidate. The alternative is to consider multiple size fractions, which necessitates a numerical approach. Width-averaged ‘sediment routing models’ that predict the coevolution of bed elevation and GSD have been applied to a range of disequilibrium situations (e.g., Talbot and Lapointe, 2002; Cui and Parker, 2005; Ferguson et al., 2006) and show encouraging agreement with ﬁeld evidence. If set up with straight long proﬁles (but possibly unsteady discharge) they could be used to investigate regime as well as transient change. The main limitations of the current generation of models are width averaging, which may lead to underestimation of bedload transport (see Section 2.3), and the assumption of no change in width during aggradation or degradation. Allowance for these complications would require either going over to a full 2-D model (see Section 4) or parameterisation in what might be thought of as a 1.5-D model (e.g., the allowance for width change in Carroll et al., 2004).

2.3.

The significance of spatial and temporal variability

As noted in Section 2.1, traditional regime approaches to large-scale self-organisation assume uniform channel geometry and a dominant discharge, whereas natural channels are spatially non-uniform because of within-reach self-organisation and experience a spectrum of ﬂoods of different magnitude, duration, and interval. Spatial and temporal variability has major implications for in-stream ecology, and affects the realism of calculations in analytical regime theory and 1-D numerical models. Many engineered channels are straight with constant width and trapezoidal cross section. This gives uniform ﬂow depth, and near-uniform velocity and shear stress, across and along the ﬂat bed. A natural gravel-bed reach conveying the same discharge normally has a bar-pool-rifﬂe morphology with a wider range of depth and more varied combinations of depth and velocity. The ecological beneﬁts of greater diversity in physical habitat are well known, indeed they are central to modern practice in river restoration. Hydraulic diversity increases the range of ecological habitat in a reach at a given discharge and ensures that the range is maintained in ﬂood conditions: areas that become too fast-ﬂowing are replaced by newly inundated areas of shallower, slower ﬂow (e.g., Mosley, 1983). Non-uniformity in ﬂow is often accompanied by sediment sorting which further increases habitat diversity. Detailed information on the within-reach variability of ecologically relevant ﬂow variables such as depth, velocity, and shear stress is very time-consuming to obtain using traditional point measurements. Some researchers have attempted to identify transferable statistical models for the frequency distributions of these variables (e.g., Lamouroux et al., 1995; Stewardson and McMahon, 2002) and to relate distribution parameters to discharge and reach-averaged morphological properties. It is also becoming possible to obtain distributed measurements more efﬁciently using new

R. Ferguson

40

techniques (Section 3), or to simulate the 2-D or 3-D velocity ﬁeld numerically (Section 4). Until recently the emphasis in reach-scale riverine ecology has been on ﬁsh habitat, but attention has now extended to the full range of ﬂora and fauna. This introduces a further consideration: the less mobile the organism, the more important becomes temporal change in local physical habitat through ﬂoods, erosion, and deposition. Much of this is played out at sub-reach scale but it can affect reach properties. For example plant colonisation is constrained not only by spatial variability in bed elevation and GSD but also by the intervals between, and seasonal timing of, ﬂow extremes which rip plants out or dry their roots. Successful colonisation feeds back to channel hydraulics through increased roughness, and to bed GSD and bank stability through ﬁne sedimentation. Traditional regime-type characterisation of rivers in a spatially and temporally averaged way cannot begin to provide understanding of such interactions. The implications of spatial variability for regime theory and 1-D channel-change models relate to calculations of ﬂow resistance and bedload transport. Using a grain size to parameterise ﬂow resistance implies that resistance is primarily due to grain roughness. This ignores the possibly large contribution of form resistance in nonuniform channels. Ways round this are to inﬂate the resistance coefﬁcient (but by how much?) or use a 2-D model (Section 4). Calculating a river’s total bedload ﬂux from the width-averaged shear stress based on mean depth, when t in fact varies across the channel, tends to underestimate the true ﬂux because transport increases non-linearly with local t. Paola (1996) and Nicholas (2000) estimated that the concentration of ﬂow and transport in the conﬂuences of a braided river could increase bedload ﬂux to three times what a uniform channel would convey, and Ferguson (2003) calculated that even bigger differences are possible in highly non-uniform situations with ﬂow not far above threshold. Such effects are, however, offset to some extent by the development of bed patchiness. Size-selective transport leads to withinreach sediment sorting, with a ﬁner bed in locations of weaker ﬂow such as bar tails. This can almost eliminate the effect of spatial ﬂow variance on bedload conveyance (Ferguson, 2003), though bed patchiness does enhance the selectivity of transport along a reach (Paola and Seal, 1995). It may be possible to use measures of crosssection shape to parameterise spatial variance in t and allow for it in a 1.5-D model, but it is probably better to turn to 2-D modelling if channels are highly non-uniform.

3.

Quantifying within-reach variability

Until quite recently the only spatially distributed information about rivers that was readily available was planimetric: the water and vegetation margins shown by aerial photographs. Quantifying bed elevation required some form of surveying, usually levelling along cross sections. Levelled sections show lateral variation in depth well, and deﬁne mean bed elevation precisely enough to detect aggradation or degradation in a resurvey, but they show longitudinal variation in morphology far less well unless the sections are very closely spaced. Information on velocity has also normally been collected along cross sections, one point at a time. It is possible to obtain sparse

Gravel-bed rivers at the reach scale

41

distributed measurements of current speed at one or two heights, as in standard protocols for assessing physical habitat for in-stream ecology, but detailed investigation of the ﬂow ﬁeld is very laborious and few rivers remain at near-constant stage for long enough. Bed GSD and structure are directly measurable only in small areas. Obtaining gravel GSDs by bulk sieving is notoriously time consuming because large sample sizes are required for reasonable precision, so mapping spatial variation by bulk sampling is inconceivable. Pebble counting is much quicker, but attention to precision is again necessary and detailed mapping of spatial variation is a daunting task. It is fortunate, therefore, that technological developments over the last decade or so have greatly extended the possibilities for obtaining spatially distributed data about river reaches. Most of them involve remote sensing of some kind. I discuss in turn new ways to quantify river bed topography, bed material, ﬂow ﬁeld, and sediment transport. It should be noted that some of these tools have been applied to only one or a few rivers so far; they are still under development and may not be useful across the full range of gravel-bed conditions. 3.1.

Topography

Several alternatives to levelling allow the capture of more continuously distributed elevation data. On exposed gravel bars and in shallow channels, self-tracking motorised total stations allow rapid mapping of large numbers of points in a fairly random pattern for subsequent construction of a digital elevation model (DEM). This can also be done by differential GPS though with lower precision. Differencing of DEMs from successive surveys gives a fuller picture of channel change than is possible from cross sections (e.g., Brasington et al., 2003). Dense DEMs can also be constructed by digital photogrammetry using airborne or oblique terrestrial imagery (Chandler et al., 2002; Lane et al., 2003). It is possible to obtain dm-scale spatial resolution and cm-scale vertical resolution over areas of several hectares, even from non-metric imagery. When oblique imagery is used the bed topography of active channels has to be surveyed separately and merged with the photogrammetric data, but with near-vertical imagery and shallow clear water it is possible to correct for refraction errors (Westaway et al., 2001). Fig. 2.1 shows the kind of detail of channel change that can be obtained. Water depth can also be mapped from differences in radiance in multi-spectral airborne or satellite imagery (Winterbottom and Gilvear, 1997; Legleiter et al., 2004). Another recent development is the use of airborne laser surveying (LiDAR) to obtain high-resolution transects of ﬂoodplain and bar topography, even in the presence of vegetation. At the opposite extreme Butler et al. (2002) used digital photogrammetry to obtain mm-scale DEMs of bed microtopography, and Carbonneau et al. (2003) showed that even a low-cost version of this approach could deliver sub-cm precision on dry surfaces. 3.2.

Bed material

There were several attempts in the 1970s and 80s to estimate bar-surface GSDs by manual ‘photo sieving’ of close-up photographs. This saved time in the ﬁeld but

R. Ferguson

42

Distance cross-stream (m)

0

1.5

200

1

400

0.5

600

0

800

-0.5

1000 1200

-1

500

1000

1500

2000

2500

-1.5

Distance downstream (m)

Figure 2.1. Channel change in a braided reach of the Waimakariri River, New Zealand, over a 3-month period as indicated by the difference between two DEMs based on oblique digital photogrammetry. Greyscale shows difference in elevation (in metres), with deposition positive and erosion negative. (Modiﬁed from Fig. 2.2 in Lane et al. (2003). Copyright John Wiley & Sons Limited. Reproduced with permission.)

required laborious image measurements and calculations afterwards. The advent of scanners, digital cameras, powerful personal computers, and image-analysis software has allowed the development of automated procedures (Butler et al., 2001; Sime and Ferguson, 2003; Graham et al., 2005) for GSDs in areas of 1 m2, aiding fuller characterisation of sedimentary facies and habitat patches. In principle automated grain sizing could be done with underwater photographs, though this remains to be tested. The limitation of these methods is that they are restricted to the visible surface area of grains larger than a cutoff of several mm. A very promising alternative approach uses a different kind of image processing on smaller-scale imagery obtained by low-level aerial photography. Carbonneau et al. (2004) successfully mapped median grain size along an 80-km reach from the local semivariance of digital images with a resolution of 3 cm (see Fig. 2.2 for a small extract), and work is under way to extend this to fuller information about surface GSD.

3.3.

Flow field

The development in the 1990s of small current meters using acoustic Doppler velocimetry (ADV) made it possible to measure three orthogonal components of instantaneous velocity. Quite apart from allowing much more detailed studies of turbulence, ADV current meters have been used to map ﬂow structures in conﬂuences and bends in more detail than before (e.g., Frothingham and Rhoads, 2003). ADV data give direct evidence on up and downwelling which previously was inferred from cross-stream components. An alternative strategy is to use ADV measurements to validate a 3-D ﬂow model and use that to inspect the ﬂow structure (Ferguson et al., 2003).

Gravel-bed rivers at the reach scale

43

Figure 2.2. Aerial photo (left, with 20-m scale bar) of a bar in the Rivie`re Sainte Marguerite, Que´bec, Canada and map (right) of median grain size (in cm) on the same bar as derived by image processing using the method of Carbonneau et al. (2004). Note correct detection of coarse/ﬁne alternation across the bar tail, and of low vegetation.

Acoustic Doppler current proﬁling (aDcp) extends ADV technology to measure instantaneous velocity components simultaneously at many heights. The strong return from the bed provides the local ﬂow depth. The near-surface ﬂow cannot be measured, and near-bed values are unreliable because of sidelobe interference, but if enough intermediate points are sampled it is possible to get a precise estimate of the depth-averaged velocity. The deployment of aDcp from a moving boat, with GPS positioning, is becoming the method of choice for discharge measurement in large rivers, and the potential uses of aDcp are far wider than this. A shallow-water aDcp instrument can be deployed from a towed or tethered raft in a small river to capture data far faster than by ADV. This should allow ﬂow ﬁelds to be mapped in far more detail than was previously possible. Also, since aDcp gives the velocity proﬁle, the bed shear stress can be estimated by ﬁtting the law of the wall to part or all of the depth (Fig. 2.3).

3.4.

Sediment transport

Bed-material transport is a key process in gravel-bed rivers because it provides the feedback from ﬂow to morphology. It is highly variable in space and time because of its non-linear dependence on ﬂow, and is by far the most difﬁcult aspect of the whole ﬂuvial system to quantify reliably in the ﬁeld. Point measurement using samplers lowered onto the bed suffers from poor repeatability and a tendency to under-catch unless the sampler oriﬁce is large compared to both transported and immobile clasts. A fairly reliable estimate of the total ﬂux can be obtained by sampling for several minutes at each of many positions across a river, but spatial patterns cannot be mapped accurately. Pit traps and vortex-tube extractors give more reliable results at single points, and valuable insight into variation over time, but they have high installation and operating costs, are restricted to small channels, and give no information on spatial differences within a reach. Three promising alternatives exist: morphological estimation of ﬂuxes from volumes of erosion and deposition between channel resurveys (reviewed by Ashmore

R. Ferguson

44

Figure 2.3. Section across Lower Fraser River, western Canada, showing lateral variability in hydraulic properties as shown by moving-boat aDcp measurements at a mean spacing of about 1 m. Dashed grey line shows local depth. Solid line shows local vertically average velocity with open symbols for averages over successive 20-m bins. Dotted line shows local shear stress estimated from mean velocity with solid symbols for 20-m averages.

and Church, 1998); the use of tracer pebbles to estimate ﬂux as the product of virtual velocity and active width and depth (Haschenburger and Church, 1998) and inferring local bedload velocity from the discrepancy between apparent boat velocity from aDcp bottom tracking and true boat velocity from Differential Global Positioning System (DGPS) (Rennie et al., 2002). These methods yield information at different time and space scales. The morphological method has been applied to channels of all sizes over intervals from a day to several years. It has several variants, mostly giving only a channel-wide ﬂux and all prone to underestimation if there has been any temporal alternation of scour and ﬁll. The results are also sensitive to survey error, but it is possible to quantify this and assess necessary point densities (e.g., Brasington et al., 2003). The tracer method is feasible only in fairly small channels and estimates mean bedload ﬂux through a reach over a single ﬂood event or longer period. There is some indication that estimates over different time periods are inconsistent (Ferguson and Hoey, 2002), but this may be because the best way to estimate active depth has not yet been identiﬁed. A thorough methodological investigation at a site with a good pit or vortex-tube trap would be useful, extending the work of Sear et al. (2003). The aDcp method has the greatest uncertainties in arriving at a quantitative estimate of ﬂux, but also the greatest potential to reveal spatial and temporal differences in the intensity of bed movement. 4.

Spatially distributed modelling of gravel-bed rivers

If attention is restricted to uniform channels and ﬂow, the implications of assumptions about process can often be deduced mathematically. This is seldom possible for

Gravel-bed rivers at the reach scale

45

non-uniform situations (an exception is the analysis by Repetto et al. (2002) of how bars are forced by width variation) so it is necessary to resort to numerical solutions. Fortunately the technological advances which have enabled spatially distributed monitoring also permit distributed numerical modelling, with large numbers of computational cells across as well as along a reach (a 2-D model), and perhaps also in the vertical (a 3-D model). This gives the opportunity to compare theoretically based predictions with observed river behaviour and distributed measurements, and to use models to simulate unobservable situations or what-if scenarios. Distributed models do however require more input data and user expertise. They vary in scope, process representation, grid type, and numerical solution method so almost every model is different. The literature is expanding rapidly and it is impossible to attempt more than a brief overview here; the chapters in Bates et al. (2005) contain detailed reviews of different aspects. A model may predict just the ﬂow ﬁeld, or ﬂow together with sediment transport, or possibly also channel change. In the ﬁrst two cases the channel is a ﬁxed container represented by a grid. In morphological models the bed level, and perhaps also bed GSD, of each cell is updated at intervals; some models also simulate bank erosion, which requires remeshing of the grid.

4.1.

Flow models

The fundamental component of all models is a ﬂow solver which computes the velocity ﬁeld over the present channel conﬁguration. There are three broad categories: 3-D computational ﬂuid dynamics (CFD), 2-D CFD, and simpliﬁed 2-D. CFD models solve discretized versions of the 3-D Navier–Stokes equations for conservation of ﬂuid mass and momentum, or the 2-D St. Venant equations which are the depth-averaged equivalent. Standard 3-D CFD codes treat momentum transfer by turbulence in a statistical way using any of several sub-models for the production and transport of turbulent kinetic energy. Vertical and lateral shear is explicitly modelled except in the cells touching the bed and banks where velocity is computed from the law of the wall with a speciﬁed roughness height. A rectangular computational grid is distorted into boundary-ﬁtted coordinates (BFC) which approximate the 3-D channel geometry, and the equations are solved by ﬁnite-volume methods. This kind of model has been applied to short reaches of natural rivers by Sinha et al. (1998), Booker et al. (2001), and others. A promising alternative to BFC uses a rectangular grid with irregular topography represented by blocking cells out (Olsen and Stokseth, 1995). This permits explicit modelling of the effects of bed microtopography (Lane et al., 2004) and should eventually be computationally feasible at reach scale. In 2-D CFD, which at present is all that is possible when modelling long reaches, the St. Venant equations are solved on a BFC grid or triangular ﬁnite-element mesh. Lateral shear is represented by an eddy viscosity. Some models include a representation of how curvature-induced secondary circulation affects the primary ﬂow. Bed roughness is parameterised, usually by Manning’s n. Leclerc et al. (1995) and Crowder and Diplas (2000) advocated such models for ecological applications, and Lisle et al. (2000) demonstrated the potential in geomorphology. Fig. 2.4 gives an

R. Ferguson

46

Figure 2.4. Extract from a 2-D ﬂow simulation of a 12-km reach of Lower Fraser River, Canada, at 60% of mean annual ﬂood, made using the DHI MIKE21C code. Area shown is about 2.5 1.5 km (axes are labelled in UTM metres). Depth is shown by shading, vertically averaged velocity by arrows (maximum speed about 3 m s 1). Note ﬂow divergence around bar head and recirculation in lee.

example of the kind of detail that can be simulated within a long non-uniform reach. A 2-D model that contains an allowance for secondary-circulation effects can successfully simulate the line of the fastest current even in bends, and the existence and location of recirculation. Finally, a few reach- or basin-scale morphological models differ radically by using a square or hexagonal grid and simpliﬁed ﬂow routing rules which conserve mass but not momentum (Murray and Paola, 1994; Coulthard et al., 2002). These models do not maintain a smooth water surface and the ﬂow ﬁeld depends not just on the routing rules but also the type and resolution of the grid, in contrast to CFD where the aim is to achieve grid-independent results (Nicholas, 2005).

4.2.

Transport and morphology

A distributed ﬂow model can drive calculations of bedload transport, and it is then straightforward to compute changes in bed level from the divergence of the sediment transport vector. In CFD-based models the driver for transport is t, calculated from velocity squared, but the simpliﬁed rule-based models use unit discharge and bed slope. The ﬁrst 2-D morphological models assumed uniform sediment but a few now allow multiple sizes, as in the 1-D models mentioned in Sections 2.1 and 2.2. Some also allow for the effect of local bed slope on transport direction, which is important in multi-size calculations. A few incorporate simple algorithms for bank erosion and thus planform change, though this is currently the weakest part of these models

Gravel-bed rivers at the reach scale

47

(Mosselman, 1998). Simulating bedload transport requires assumptions about sediment ﬂux into the reach, and (for multi-size calculations) the GSDs of this ﬂux and of the initial bed in each cell. Packages of this type have been developed and extended over many years by Delft Hydraulics (Netherlands) and DHI (Denmark) for use in engineering consultancy projects, and models have also been developed by the U.S. Geological Survey (building on Nelson and Smith, 1989) and a growing number of university engineering departments. Little of the commercially directed development is reported in the periodical literature so it is not surprising that there has been some ‘reinventing the wheel’ by academic researchers (Mosselman, 2004). Recent published developments, especially as regards simulating planform change, are discussed by Lane and Ferguson (2005). There have been very few applications of 2-D bedload or morphological modelling to gravel-bed rivers so far, but there is vast potential whether for tackling practical issues of aggradation and degradation or for investigating generic behaviour such as the self-organisation of bars, bends, and braids. Fig. 2.5 gives an example from the work I am doing on gravel transport and bar growth in the lower Fraser River. The striking feature, impossible to observe using available methods or to model in 1-D, is

Figure 2.5. Extract from a 2-D bedload transport simulation for a 12-km reach of Lower Fraser River at 60% of mean annual ﬂood. Area shown is about 3.5 2.5 km (axes are labelled in UTM metres). Flow enters at top right past a headland. Bed elevation is shown by shading and bedload transport rate by arrows.

R. Ferguson

48

that high ﬂuxes occur only locally, and on bar ﬂanks rather than in talwegs. This accords with the observed locations and styles of bar growth.

5.

Looking ahead

Scientiﬁc understanding advances fastest through a critical interplay between observation and theory. In the context of gravel-bed rivers, the tensions and inconsistencies between traditionally separate perspectives (Section 2) point to a need to move beyond regime theory, and to learn more about within-reach variability and its implications. Powerful new techniques for spatially distributed observation (Section 3) and modelling (Section 4) should be valuable tools in tackling these agendas, but how do we know we can trust them? In this concluding discussion I consider the two research agendas ﬁrst, then the methodological issues. Regime theory may be reaching its inherent limits. Eaton et al. (2004) showed there are no unique solutions even after making an optimisation assumption: just inﬁnite combinations of three dimensionless variables. Planform is not adequately linked in, and key simplifying assumptions of the whole approach (dominant discharge, uniform channels, ﬁxed grain size) are doubtful. Not all reaches are in regime and many practical problems relate to disequilibrium. Moving beyond the regime approach requires greater integration of work relating to different degrees of freedom in channel change: bed elevation and channel slope, bed GSD, and lateral adjustment of width and planform. 1-D sediment routing models link the ﬁrst two but do not yet contain versatile parameterisations of width adjustment, and are intrinsically inadequate for modelling strongly meandering or braided reaches. Phenomenological studies are needed of how width and planform alter during incision or aggradation in a range of circumstances, to see if any generalisations can be made. 2-D morphodynamic modelling avoids the limitations of 1-D models and appears to have great potential for studying within-reach self-organisation, but requires thorough testing and sensitivity analysis before widespread adoption. Emphasis on the ecological as well as engineering health of rivers highlights the importance of spatial variability and temporal ﬂuctuation. Probability-distribution models that extend the hydraulic geometry concept deserve further attention, focused on testing transferability between reaches and predictability of distribution parameters. I suspect distributed ﬂow modelling is a more universally applicable approach since there is a big literature on the general validity of the approximations made and it is becoming easier to obtain detailed reach-speciﬁc test data. Open questions include whether 3-D rather than 2-D ﬂow detail is needed for some ecological purposes, and how to include vegetation dynamics in a reach-scale ﬂow (or morphodynamic) model. The ﬁnal set of issues concerns the interplay of observation and theory, the latter increasingly in the form of models. To advocate wider use of numerical modelling is not to suggest it can replace ﬁeld investigations and ﬂume experiments. Measurements are necessary to set up a model of any particular situation, and they are traditionally the test of a model: any serious discrepancy between model predictions and real-world observations is taken to cast doubt on the process representation in

Gravel-bed rivers at the reach scale

49

the model, though in a complicated model it may not be obvious where the problem lies. In fact the difference between predictions and observations depends on far more than process representation (Fig. 2.6). Real-world causal mechanisms act in particular circumstances to produce outcomes, but those outcomes depend on our decision to study one system rather than another: the experimental control or site selection that excludes some processes or factors and thereby emphasises others (Richards, 1990). Moreover, the ‘observations’ we compare with predictions are a ﬁltered and REAL WORLD ‘real’ mechanisms contingent circumstances

site selection or experimental design

outcomes in specific situation OBSERVER

observational methods

observations

NUMERICAL MODEL

assumptions

process representation

initial and boundary conditions MODELLER

model predictions

resolution and numerics

Figure 2.6. Sources of ambiguity when comparing predictions from a numerical model with observations from a river. Upper part of diagram shows real-world situation, lower part shows model. Human decisions are involved in both observation and modelling.

50

R. Ferguson

blurred sample of the spatially and temporally continuous variables of interest. What we observe, at what spatial and temporal resolution, is a subjective decision constrained by available technology and resources. All measurement is imprecise, to a degree that depends on choices of technique, instrument, quality control, and data processing. Finally, model predictions are themselves a blurred representation of the consequences of the process assumptions, ﬁltered through the grid design and numerical solution methods. Thus two attempts to model the same situation and test against measurements could come up with different predictions, observations, agreement, and conclusions. This ambiguity can be reduced by several strategies. Numerics, gridding and the representation of speciﬁc processes can be tested in simple situations with highquality measurements (e.g., ﬂume experiments). Opportunities can be taken to validate internal variables in more physically based models (e.g., depth and velocity in morphodynamic models where the main interest is in transport and channel change), though not in rule-based models that claim only to capture key qualitative aspects of high-level behaviour. Quantitative tests of spatially distributed models can be complemented by qualitative assessment of agreement in spatial patterns. But there will always be scope for disagreement about the adequacy of a model, not just because of the ambiguities inherent in testing but also because ‘adequacy’ has to be relative to purpose and the resources available. One response to the complexity of natural rivers is to seek ever ﬁner spatial and temporal resolution, and represent smallscale processes explicitly rather than ignoring or parameterising them, but this reductionist strategy will never completely displace simpler models which can be applied to longer reaches with less information, or to generic situations. The most important role of cutting-edge models may instead be to establish the conditions in which simpliﬁcations are acceptable, and help develop better parameterisations for simpler models. These are exciting times for reach-scale study of gravel-bed rivers, and we need to take full advantage of the opportunities afforded by technical developments to help ﬁll the gaps in understanding that are shown by comparing regime, channel-change, bedload, and ecological perspectives. This requires an interplay of empirical and theoretical approaches. Spatially distributed modelling has as much of a role to play as spatially distributed monitoring, but both need further development and testing before they can become routine tools. We need to consider critically the reliability of both observations and predictions, and to compare models of different complexity to decide what simpliﬁcations are adequate for what purposes.

Acknowledgements I thank my colleagues Stuart Lane and Patrice Carbonneau for Figs. 2.1 and 2.2 respectively. The Fraser River data illustrated in Fig. 2.3 were obtained under Canadian NSERC grant 246057 to Mike Church; Louise Sime processed the ADCP data. The modelling results illustrated in Figs. 2.4 and 2.5 were obtained as part of UK NERC grant NER/B/S/2002/00354. The Royal Society supported my participation in the GBR6 workshop.

Gravel-bed rivers at the reach scale

51

References Ashmore, P., Church, M., 1998. Sediment transport and river morphology: A paradigm for study. In: Klingeman, P.C., Beschta, R.L., Komar, P.D., and Bradley, J.B. (Eds), Gravel-Bed Rivers in the Environment. Water Resource Publications, Colorado, pp. 115–148. Ashmore, P.E., 1991. How do gravel-bed rivers braid? Can. J. Earth Sci. 28, 326–341. Bates, P.D., Lane, S.N., Ferguson, R.I. (Eds), 2005. Computational ﬂuid dynamics: Applications in environmental hydraulics. Wiley, Chichester, 531pp. Booker, D.J., Sear, D.A., Payne, A.J., 2001. Modelling three-dimensional ﬂow structures and patterns of boundary shear stress in a natural pool-rifﬂe sequence. Earth Surf. Process. Landf. 26, 553–576. Brasington, J., Langham, J., Rumsby, B., 2003. Methodological sensitivity of morphometric estimates of coarse ﬂuvial sediment transport. Geomorphology 53, 299–316. Bufﬁngton, J.M., Montgomery, D.R., 1999. Effects of sediment supply on surface textures of gravel-bed rivers. Water Resour. Res. 35, 3523–3530. Butler, J.B., Lane, S.N., Chandler, J.H., 2001. Automated extraction of grain-size data from gravel surfaces using digital image processing. J. Hydraul. Res. 39, 519–529. Butler, J.B., Lane, S.N., Chandler, J.H., Porﬁri, K., 2002. Through-water close range digital photogrammetry in ﬂume and ﬁeld environments. Photogramm. Rec. 17, 419–439. Carbonneau, P.E., Lane, S.N., Bergeron, N., 2003. Cost-effective non-metric close-range digital photogrammetry and its application to a study of coarse gravel river beds. Int. J. Remote Sens. 24, 2837–2854. Carbonneau, P.E., Lane, S.N., Bergeron, N., 2004. Catchment-scale mapping of surface grain size in gravel bed rivers using airborne digital imagery. Water Resour. Res. 40, art. no. W07202. Carroll, R.W.H., Warwick, J.J., James, A.I., Miller, J.R., 2004. Modeling erosion and overbank deposition during extreme ﬂood conditions on the Carson River, Nevada. J. Hydrol. 297, 1–21. Chandler, J., Ashmore, P., Paola, C., et al., 2002. Monitoring river channel change using terrestrial oblique digital imagery and automated digital photogrammetry. Ann. Assoc. Am. Geographers 92, 631–644. Church, M., Hassan, M.A., Wolcott, J.F., 1998. Stabilizing self-organized structures in gravel-bed stream channels: Field and experimental observations. Water Resour. Res. 34, 3169–3180. Coulthard, T.J., Macklin, M.G., Kirkby, M.J., 2002. A cellular model of Holocene upland river basin and alluvial fan evolution. Earth Surf. Process. Landf. 27, 269–288. Crowder, D.W., Diplas, P., 2000. Using two-dimensional hydrodynamic models at scales of ecological importance. J. Hydrol. 230, 172–191. Cui, Y.T., Parker, G., 2005. Numerical model of sediment pulses and sediment-supply disturbances in mountain rivers. J. Hydr. Eng. ASCE 131, 646–656. Darby, S.E., et al., 1998. River width adjustment. II: Modeling (task force report). J. Hydr. Eng. ASCE 124, 903–917. Deigaard, R., Fredsoe, J., 1978. Longitudinal grain sorting by current in alluvial streams. Nord. Hydrol. 9, 7–16. Dietrich, W.E., Kirchner, J.W., Ikeda, H., Iseya, F., 1989. Sediment supply and the development of the coarse surface-layer in gravel-bedded rivers. Nature 340, 215–217. Eaton, B.C., Church, M., Millar, R.G., 2004. Rational regime model of alluvial channel morphology and response. Earth Surf. Process. Landf. 29, 511–529. Ferguson, R., Hoey, T., Wathen, S., Werritty, A., 1996. Field evidence for rapid downstream ﬁning of river gravels through selective transport. Geology 24, 179–182. Ferguson, R.I., 2003. The missing dimension: Effects of lateral variation on 1-D calculations of ﬂuvial bedload transport. Geomorphology 56, 1–14. Ferguson, R.I., Cudden, J.R., Hoey, T.B., Rice, S.P., 2006. River system discontinuities due to lateral inputs: Generic styles and controls. Earth Surf. Proc. Landf. 31, 1149–1166. Ferguson, R.I., Hoey, T.B., 2002. Long-term slowdown of river tracer pebbles: Generic models and implications for interpreting short-term tracer studies. Water Resour. Res. 38, art. no. 1142. Ferguson, R.I., Parsons, D.R., Lane, S.N., Hardy, R.J., 2003. Flow in meander bends with recirculation at the inner bank. Water Resour. Res. 39, art. no. 1322.

52

R. Ferguson

Fonstad, M., Marcus, M.A., 2003. Self-organized criticality in riverbank systems. Ann. Assoc. Am. Geographers 93, 281–296. Frothingham, K.M., Rhoads, B.L., 2003. Three-dimensional ﬂow structure and channel change in an asymmetrical compound meander loop, Embarras River, Illinois. Earth Surf. Process. Landf. 28, 625–644. Graham, D.J., Reid, I., Rice, S.P., 2005. Automated sizing of coarse-grained sediments: Image-processing procedures. Math. Geol. 37, 1–28. Haschenburger, J.K., Church, M., 1998. Bed material transport estimated from the virtual velocity of sediment. Earth Surf. Process. Landf. 23, 791–808. Lamouroux, N., Souchon, Y., Herouin, E., 1995. Predicting velocity frequency-distributions in stream reaches. Water Resour. Res. 31, 2367–2375. Lane, E.W., 1955. The importance of ﬂuvial morphology in river hydraulic engineering. Proc. Am. Soc. Civil Eng. 81, 1–17. Lane, S.N., Ferguson, R.I., 2005. Modelling reach-scale ﬂuvial ﬂows. In: Bates, P.D., Lane, S.N., and Ferguson, R.I. (Eds), Computational Fluid Dynamics: Applications in Environmental Hydraulics. Wiley, Chichester, pp. 217–269. Lane, S.N., Hardy, R.J., Ingham, D.B., Elliott, L., 2004. Numerical modeling of ﬂow processes over gravelly surfaces using structured grids and a numerical porosity treatment. Water Resour. Res. 40, art. no. W01302. Lane, S.N., Westaway, R.M., Hicks, D.M., 2003. Estimation of erosion and deposition volumes in a large, gravel-bed, braided river using synoptic remote sensing. Earth Surf. Process. Landf. 28, 249–271. Leclerc, M., Boudreault, A., Bechara, J.A., Corfa, G., 1995. Two-dimensional hydrodynamic modeling – a neglected tool in the instream ﬂow incremental methodology. Trans. Am. Fish. Soc. 124, 645–662. Legleiter, C.J., Roberts, D.A., Marcus, W.A., Fonstad, M.A., 2004. Passive optical remote sensing of river channel morphology and in-stream habitat: Physical basis and feasibility. Remote Sens. Environ. 93, 493–510. Leopold, L.B., Maddock, T., 1953. The hydraulic geometry of stream channels and some physiographic implications. U.S. Geol. Surv. Prof. Paper 252, 64. Lewin, J., Macklin, M.G., Newson, M.D., 1988. Regime theory and environmental change – irreconcilable concepts? In: White, W.R. (Ed.), International Conference on River Regime. Wiley, pp. 431–445. Lie´bault, F., Pie´gay, H., 2002. Causes of 20th century channel narrowing in mountain and Piedmont rivers of southeastern France. Earth Surf. Process. Landf. 27, 425–444. Lisle, T.E., Nelson, J.M., Pitlick, J., et al., 2000. Variability of bed mobility in natural, gravel-bed channels and adjustments to sediment load at local and reach scales. Water Resour. Res. 36, 3743–3755. McLean, D.G., Church, M., 1999. Sediment transport along lower Fraser River – 2. Estimates based on the long-term gravel budget. Water Resour. Res. 35, 2549–2559. Medawar, P.B., 1967. The Art of the Soluble. Methuen. (Also incorporated in Medawar, P.B., 1982, Pluto’s Republic, Oxford University Press, 351pp.) Millar, R.G., Quick, M.C., 1998. Stable width and depth of gravel-bed rivers with cohesive banks. J. Hydraul. Eng. ASCE 124, 1005–1013. Montgomery, D.R., Panﬁl, M.S., Hayes, S.K., 1999. Channel-bed mobility response to extreme sediment loading at Mount Pinatubo. Geology 27, 271–274. Mosley, M.P., 1983. Response of braided rivers to changing discharge. NZ J. Hydrol. 22, 18–67. Mosselman, E., 1998. Morphological modelling of rivers with erodible banks. Hydrol. Process. 12, 1357–1370. Mosselman, E., 2004. Discussion of ‘numerical modeling of bed evolution in channel bends’ by Kassem, A.A., Chaudhry, M.H,. J. Hydraul. Eng. ASCE 130, 82. Murray, A.B., Paola, C., 1994. A cellular model of braided rivers. Nature 371, 54–57. Nelson, J.M., Smith, J.D., 1989. Evolution and stability of erodible channel beds. In: Ikeda, S. and Parker, G. (Eds), River Meandering, AGU Geophys. Monogr., pp. 321–378. Nicholas, A.P., 2000. Modelling bedload yield in braided gravel bed rivers. Geomorphology 36, 89–106.

Gravel-bed rivers at the reach scale

53

Nicholas, A.P., 2005. Cellular modelling in ﬂuvial geomorphology. Earth Surf. Process. Landf. 30, 645–649. Olsen, N.R.B., Stokseth, S., 1995. Three-dimensional numerical modeling of water-ﬂow in a river with large bed roughness. J. Hydraul. Res. 33, 571–581. Paola, C., 1996. Incoherent structure: Turbulence as a metaphor for stream braiding. In: Ashworth, P.J., Bennett, S.J., Best, J.L., and McLelland, S.J. (Eds), Coherent Flow Structures in Open Channels. Wiley, Chichester, pp. 705–723. Paola, C., Seal, R., 1995. Grain size patchiness as a cause of selective deposition and downstream ﬁning. Water Resour. Res. 31, 1395–1407. Parker, G., 1979. Hydraulic geometry of active gravel rivers. J. Hydraulics Div. ASCE 105, 1185–1201. Parker, G., Klingeman, P.C., 1982. On why gravel bed streams are paved. Water Resour. Res. 18, 1409–1423. Pitlick, J., 1993. Response and recovery of a subalpine stream following a catastrophic ﬂood. Geol. Soc. Am. Bull. 105, 657–670. Rennie, C.D., Millar, R.G., Church, M., 2002. Measurement of bed load velocity using an acoustic Doppler current proﬁler. J. Hydraul. Eng. ASCE 128, 473–483. Repetto, R., Tubino, M., Paola, C., 2002. Planimetric instability of channels with variable width. J. Fluid Mech. 457, 79–109. Rhoads, B.L., Welford, M.R., 1991. Initiation of river meandering. Prog. Phys. Geogr. 15, 127–156. Richards, K., 1990. Real geomorphology. Earth Surf. Process. Landf. 15, 195–197. Rinaldi, M., 2003. Recent channel adjustments in alluvial rivers of Tuscany, central Italy. Earth Surf. Process. Landf. 28, 587–608. Schumm, S.A., 1969. River metamorphosis. J. Hydr. Div. ASCE 95, 255–273. Sear, D.A., Lee, M.W.E., Carling, P.A., et al., 2003. An assessment of the accuracy of the spatial integration method (SIM) for estimating coarse bedload transport in gravel-bedded streams using tracers. Int. Assoc. Hydrol. Sci. Publ. 283, 164–171. Sime, L.C., Ferguson, R.I., 2003. Information on grain sizes in gravel-bed rivers by automated image analysis. J. Sediment. Res. 73, 630–636. Sinha, S.K., Sotoropoulos, F., Odgaard, A.J., 1998. Three-dimensional numerical model for ﬂow through natural rivers. J. Hydraul. Eng. ASCE 124, 13–24. Stewardson, M.J., McMahon, T.A., 2002. A stochastic model of hydraulic variations within stream channels. Water Resour. Res. 38, art. no. 1007. Stolum, H.H., 1998. Planform geometry and dynamics of meandering rivers. Geol. Soc. Am. Bull. 110, 1485–1498. Talbot, T., Lapointe, M., 2002. Numerical modeling of gravel bed river response to meander straightening: The coupling between the evolution of bed pavement and long proﬁle. Water Resour. Res. 38, art. no. 1074. Van der Nat, D., Tockner, K., Edwards, P.J., et al., 2003. Habitat change in braided ﬂood plains (Tagliamento, NE Italy). Freshwater Biol. 48, 1799–1812. Ward, J.V., Tockner, K., Arscott, D.B., Claret, C., 2002. Riverine landscape diversity. Freshwater Biol. 47, 517–539. Westaway, R.M., Lane, S.N., Hicks, D.M., 2001. Remote sensing of clear-water, shallow, gravel-bed rivers using digital photogrammetry. Photogramm. Eng. Remote Sens. 67, 1271–1281. Wilcock, P.R., 2001. The ﬂow, the bed, and the transport: Interaction in ﬂume and ﬁeld. In: Mosley. M.P. (Ed.), Gravel-bed rivers V. NZ Hydrol. Soc., Wellington, pp. 183–220. Wilcock, P.R., Crowe, J.C., 2003. Surface-based transport model for mixed-size sediment. J. Hydraul. Eng. ASCE 129, 120–128. Winterbottom, S.J., Gilvear, D.J., 1997. Quantiﬁcation of channel bed morphology in gravel-bed rivers using airborne multispectral imagery and aerial photography. Regulated Rivers: Research and Management 13, 489–499. Wyzga, B., 2001. A geomorphologist’s criticism of the engineering approach to channelization of gravelbed rivers: Case study of the Raba River, Polish Carpathians. Environmental Management 28, 341–358.

54

R. Ferguson

Discussion by D. Milan, G. Heritage and D. Hetherington R. Ferguson mentions several important technological advances in the acquisition of data on river bed topography. Many of the techniques (aerial LIDAR, EDM theodolites, GPS, photogrammetry) suffer coverage or resolution limitations resulting in a trade off between spatial coverage and morphologic detail captured (Fig. 2.7). This issue is particularly important when rates of spatial and temporal change are considered for ﬂuvial systems. At a reach scale the acquisition of high-resolution topographic information is central to the effective construction of a DEM. Oblique ﬁeld based laser scanning (LiDAR) now offers a signiﬁcant improvement in the speed of data capture, accuracy/resolution and areal coverage of topographic data acquisition. A Rigel LMS Z210 scanning laser has been used at a number of sites in the UK and Switzerland by the authors to collect a series of independent data sets recording range distance, relative height, surface colour and reﬂectivity. The instrument works on the principle of ‘time of ﬂight’ measurement using a pulsed eye-safe infrared laser source (0.9 mm wavelength) emitted in precisely deﬁned angular directions controlled by a spinning mirror arrangement. A sensor records the time taken for light to be reﬂected from the incident surface. Once the scanner unit is mounted on a tripod, it is capable of scanning through 801 in the vertical and 3301 in the horizontal, stepping 0.072–0.361 depending on the resolution required and the time available for scanning. Vertical scanning rates vary between 5 and 32 scans s 1. Angular measurements are recorded to an accuracy of 0.0361 in the vertical and 0.0181 in the horizontal. Range error is 0.025 m to a radial distance of 350 m. Point densities of 10,000 m 2 have been achieved from 12 meshed scans covering and area of 2000 m2, sufﬁcient to represent the surface at the grain scale. Five million

Figure 2.7. Spatial and temporal limitations of morphological capture techniques.

Gravel-bed rivers at the reach scale

55

data points may be collected in 5 min, and a vertical resolution of around 70.02 m has been achieved. The technology works best on dry river beds and exposed bar surfaces, however where the water is clear, calm, and shallow and the angle of incidence of the laser pulse is high, some penetration does occur. The speed, coverage and accuracy of the new technology clearly offers a considerable advance for topographic data acquisition in ﬂuvial systems. Discussion by H. Lamarre and A. Roy The paper points out that the reach-scale is associated with non-uniform ﬂuvial processes that recent technological advances may help in understanding. Although we agree that these advances have enhanced our ability to measure in detail the ﬂuvial system, the applicability of several methods is still limited to speciﬁc environments. For instance, several techniques to measure bed topography or ﬂow velocity are not suitable in heavily vegetated, steep, or very coarse-grained bed channels. In this context, more traditional methods are still required and, as for new technologies, they raise recurrent and crucial unresolved questions on sampling strategies. We are wondering how consideration of reach-scale properties should be used in the optimisation of the sampling designs. When measuring at the reach-scale using performing instrumentation and adequate strategies, it is possible to capture both the local scale characteristics, such as a pebble cluster or large protruding clast as well as larger features that are characteristics of the reach-scale morphology. What is the balance between the local and the reach-scale representation? Using the maximum spatial sampling density of velocity or topography that is possible to achieve given the survey time, energy, costs and limitations of the instruments, we can capture the effects of local scale characteristics which may be less signiﬁcant for the reach-scale dynamics (e.g., Lamarre and Roy, 2005). If the sampling density is reduced, only larger features remain. Determining the level of detail that must be represented at the ‘reach-scale resolution’ is then essential to a useful integration of different scales of roughness and ﬂow structures in gravel-bed rivers. Even though the optimisation of the sampling design should also include the objectives of the study and the scales of interest, what criteria should be used to determine the resolution of reach-scale surveys? References Lamarre, H., Roy, A.G., 2005. Reach-scale variability of turbulent ﬂow characteristics in a gravel-bed river. Geomorphology 69 (1–2), 95–113.

Discussion by C.D. Rennie The author has presented a comprehensive review of the state-of-the-art in 1D and spatially distributed approaches to understanding river morphology. The paper concludes with a discussion of the interplay between monitoring and modelling of spatially distributed processes in rivers. I wholeheartedly agree that monitoring

Gravel-bed rivers at the reach scale

55

data points may be collected in 5 min, and a vertical resolution of around 70.02 m has been achieved. The technology works best on dry river beds and exposed bar surfaces, however where the water is clear, calm, and shallow and the angle of incidence of the laser pulse is high, some penetration does occur. The speed, coverage and accuracy of the new technology clearly offers a considerable advance for topographic data acquisition in ﬂuvial systems. Discussion by H. Lamarre and A. Roy The paper points out that the reach-scale is associated with non-uniform ﬂuvial processes that recent technological advances may help in understanding. Although we agree that these advances have enhanced our ability to measure in detail the ﬂuvial system, the applicability of several methods is still limited to speciﬁc environments. For instance, several techniques to measure bed topography or ﬂow velocity are not suitable in heavily vegetated, steep, or very coarse-grained bed channels. In this context, more traditional methods are still required and, as for new technologies, they raise recurrent and crucial unresolved questions on sampling strategies. We are wondering how consideration of reach-scale properties should be used in the optimisation of the sampling designs. When measuring at the reach-scale using performing instrumentation and adequate strategies, it is possible to capture both the local scale characteristics, such as a pebble cluster or large protruding clast as well as larger features that are characteristics of the reach-scale morphology. What is the balance between the local and the reach-scale representation? Using the maximum spatial sampling density of velocity or topography that is possible to achieve given the survey time, energy, costs and limitations of the instruments, we can capture the effects of local scale characteristics which may be less signiﬁcant for the reach-scale dynamics (e.g., Lamarre and Roy, 2005). If the sampling density is reduced, only larger features remain. Determining the level of detail that must be represented at the ‘reach-scale resolution’ is then essential to a useful integration of different scales of roughness and ﬂow structures in gravel-bed rivers. Even though the optimisation of the sampling design should also include the objectives of the study and the scales of interest, what criteria should be used to determine the resolution of reach-scale surveys? References Lamarre, H., Roy, A.G., 2005. Reach-scale variability of turbulent ﬂow characteristics in a gravel-bed river. Geomorphology 69 (1–2), 95–113.

Discussion by C.D. Rennie The author has presented a comprehensive review of the state-of-the-art in 1D and spatially distributed approaches to understanding river morphology. The paper concludes with a discussion of the interplay between monitoring and modelling of spatially distributed processes in rivers. I wholeheartedly agree that monitoring

Gravel-bed rivers at the reach scale

55

data points may be collected in 5 min, and a vertical resolution of around 70.02 m has been achieved. The technology works best on dry river beds and exposed bar surfaces, however where the water is clear, calm, and shallow and the angle of incidence of the laser pulse is high, some penetration does occur. The speed, coverage and accuracy of the new technology clearly offers a considerable advance for topographic data acquisition in ﬂuvial systems. Discussion by H. Lamarre and A. Roy The paper points out that the reach-scale is associated with non-uniform ﬂuvial processes that recent technological advances may help in understanding. Although we agree that these advances have enhanced our ability to measure in detail the ﬂuvial system, the applicability of several methods is still limited to speciﬁc environments. For instance, several techniques to measure bed topography or ﬂow velocity are not suitable in heavily vegetated, steep, or very coarse-grained bed channels. In this context, more traditional methods are still required and, as for new technologies, they raise recurrent and crucial unresolved questions on sampling strategies. We are wondering how consideration of reach-scale properties should be used in the optimisation of the sampling designs. When measuring at the reach-scale using performing instrumentation and adequate strategies, it is possible to capture both the local scale characteristics, such as a pebble cluster or large protruding clast as well as larger features that are characteristics of the reach-scale morphology. What is the balance between the local and the reach-scale representation? Using the maximum spatial sampling density of velocity or topography that is possible to achieve given the survey time, energy, costs and limitations of the instruments, we can capture the effects of local scale characteristics which may be less signiﬁcant for the reach-scale dynamics (e.g., Lamarre and Roy, 2005). If the sampling density is reduced, only larger features remain. Determining the level of detail that must be represented at the ‘reach-scale resolution’ is then essential to a useful integration of different scales of roughness and ﬂow structures in gravel-bed rivers. Even though the optimisation of the sampling design should also include the objectives of the study and the scales of interest, what criteria should be used to determine the resolution of reach-scale surveys? References Lamarre, H., Roy, A.G., 2005. Reach-scale variability of turbulent ﬂow characteristics in a gravel-bed river. Geomorphology 69 (1–2), 95–113.

Discussion by C.D. Rennie The author has presented a comprehensive review of the state-of-the-art in 1D and spatially distributed approaches to understanding river morphology. The paper concludes with a discussion of the interplay between monitoring and modelling of spatially distributed processes in rivers. I wholeheartedly agree that monitoring

56

R. Ferguson

programs are subjectively designed, and that all observed data contain errors and represent only a portion of the actual phenomena. However, the fact remains that 2D and 3D morphodynamic models require validation with observations. Model output is becoming increasingly detailed, including mapping of the spatial distribution of bedload transport (e.g., Fig. 2.3). However, validation and calibration of morphodynamic models is usually based on comparison of observed and predicted water levels, water velocity distributions, and, perhaps, channel change. The spatial distribution of bedload transport rate, which drives the morphodynamics yet is unreliably predicted using bedload transport formulae, is not calibrated, due to the lack of available data for calibration. In order to obtain spatially distributed bedload data, which could be used to validate morphodynamic models, I have been developing the use of acoustic Doppler current proﬁlers (aDcps) to measure bedload transport velocity (Rennie et al., 2002; Rennie and Millar, 2004; Rennie and Villard, 2004; Gaueman and Rennie, in press). As stated by the author, the technique is based on the difference in boat velocity measured by DGPS and Doppler sonar (bottom track) that is biased by bedload motion. It has already been demonstrated that spatial distributions of bedload transport can be mapped in some situations (Rennie and Millar, 2004). Calibration curves have been developed to relate the observed signal to bedload transport rate (Rennie et al., 2002), although the calibration is site speciﬁc and depends on both particle size and aDcp parameters such as operating frequency and acoustic pulse length (Rennie and Villard, 2004; Gaueman and Rennie, in press). Further research is required to determine the limitations of the technique, to specify the relation between the observed signal and actual transport rate for particular aDcps in various ﬂuvial environments, and to develop statistical methods to employ the observed bedload spatial distributions for model calibration.

References Gaueman, D., Rennie, C.D. (in press). A comparison of two ﬁeld studies of acoustic bed velocity: grain size and instrument frequency effects. In 8th Federal Interagency Sedimentation Conference. Rennie, C.D., Millar, R.G., Church, M.A., 2002. Measurement of bedload velocity using an acoustic Doppler current proﬁler. J. Hydraulic Eng. (ASCE) 128, 473–483. Rennie, C.D., Millar, R.G., 2004. Measurement of the spatial distribution of ﬂuvial bedload transport velocity in both sand and gravel. Earth Surf. Process. Landf. 29 (10), 1173–1193. Rennie, C.D., Villard, P.V., 2004. Site speciﬁcity of bedload measurement using an aDcp. J. Geophys. Res. (Earth Surf.) 109 (F3), F03003, 10.1029/2003JF000106, 29 July 2004.

Discussion by G. Williams [In response to the question of Ian Reid about the ‘black hole’ of the ﬂood event itself] The best way of ‘viewing’ a river in ﬂood is through mobile bed physical models. You can see the nature of the sediment movement and the changing channel forms.

56

R. Ferguson

programs are subjectively designed, and that all observed data contain errors and represent only a portion of the actual phenomena. However, the fact remains that 2D and 3D morphodynamic models require validation with observations. Model output is becoming increasingly detailed, including mapping of the spatial distribution of bedload transport (e.g., Fig. 2.3). However, validation and calibration of morphodynamic models is usually based on comparison of observed and predicted water levels, water velocity distributions, and, perhaps, channel change. The spatial distribution of bedload transport rate, which drives the morphodynamics yet is unreliably predicted using bedload transport formulae, is not calibrated, due to the lack of available data for calibration. In order to obtain spatially distributed bedload data, which could be used to validate morphodynamic models, I have been developing the use of acoustic Doppler current proﬁlers (aDcps) to measure bedload transport velocity (Rennie et al., 2002; Rennie and Millar, 2004; Rennie and Villard, 2004; Gaueman and Rennie, in press). As stated by the author, the technique is based on the difference in boat velocity measured by DGPS and Doppler sonar (bottom track) that is biased by bedload motion. It has already been demonstrated that spatial distributions of bedload transport can be mapped in some situations (Rennie and Millar, 2004). Calibration curves have been developed to relate the observed signal to bedload transport rate (Rennie et al., 2002), although the calibration is site speciﬁc and depends on both particle size and aDcp parameters such as operating frequency and acoustic pulse length (Rennie and Villard, 2004; Gaueman and Rennie, in press). Further research is required to determine the limitations of the technique, to specify the relation between the observed signal and actual transport rate for particular aDcps in various ﬂuvial environments, and to develop statistical methods to employ the observed bedload spatial distributions for model calibration.

References Gaueman, D., Rennie, C.D. (in press). A comparison of two ﬁeld studies of acoustic bed velocity: grain size and instrument frequency effects. In 8th Federal Interagency Sedimentation Conference. Rennie, C.D., Millar, R.G., Church, M.A., 2002. Measurement of bedload velocity using an acoustic Doppler current proﬁler. J. Hydraulic Eng. (ASCE) 128, 473–483. Rennie, C.D., Millar, R.G., 2004. Measurement of the spatial distribution of ﬂuvial bedload transport velocity in both sand and gravel. Earth Surf. Process. Landf. 29 (10), 1173–1193. Rennie, C.D., Villard, P.V., 2004. Site speciﬁcity of bedload measurement using an aDcp. J. Geophys. Res. (Earth Surf.) 109 (F3), F03003, 10.1029/2003JF000106, 29 July 2004.

Discussion by G. Williams [In response to the question of Ian Reid about the ‘black hole’ of the ﬂood event itself] The best way of ‘viewing’ a river in ﬂood is through mobile bed physical models. You can see the nature of the sediment movement and the changing channel forms.

Gravel-bed rivers at the reach scale

57

Physical models of river reaches have been used very successfully in New Zealand. Rob Ferguson did not mention them in his overview. However, we have a long way to go before numerical models will give us a good representation of the reality of river dynamics. There should still be a place for physical modelling. For coarse gravel rivers, adequate scaling is relatively easily achieved, and provided you have a large shed and pumps on hand, physical models are not that expensive.

Discussion by A. Papanicolaou The author should be complimented for addressing one of the most intriguing problems in modern sediment transport theory, viz. limitations of traditional regime theories and the need to look for an ‘outside the box’ approach by treating the bed characteristics within a reach scale as a degree of freedom characteristics. The author eloquently discusses the necessity for introducing 2-D dimensional models to describe the spatial variability within a reach scale triggered by the self organization of the bed, and the three way constant feedback process among ﬂow, banks, and bed morphology that is especially pronounced in rivers that undergo a system-wide adjustment. Another issue that the discusser, though, believes should be further discussed amongst the members of this community is the issue of sediment supply within a reach scale. Traditionally sediment supply is provided from upstream of a stream reach by assuming that is originated from instream sediment sources. Recently, lateral sediment contributions from the banks have been accounted for by introducing the 1.5-D models (Wu et al., 2004). However, sediment sources originated from hillslopes and ﬂoodplains (the uplands) have not yet been considered (e.g., Gouthard et al., same issue). Upland sediment delivery transport, in some cases, could account for up to 80% of the total lateral sediment inputs within a stream reach (McCool et al., 2000). In an effort to generate new knowledge and improved understanding of the complex interrelationships between watershed and instream parameters and the scale integrity inﬂuences on channel morphology, an integrated watershed hydrologic/ sedimentation framework for mountainous watersheds must be considered. This framework must utilize advanced analytical techniques and physically based numerical models for simulating upland (macro level) and instream (micro level) processes in an integrated fashion (Wang, 2005). As a ﬁrst attempt in this direction, Papanicolaou et al. (2003) have conducted a watershed ecosystem study in the South Fork of the Clearwater River in Idaho, USA. They identiﬁed the interdependencies between watershed and instream parameters and the clusters of parameters that primarily affect sediment supply and ultimately ﬁsh health. Papanicolaou et al. (2003) concluded the following (Fig. 2.8): (1) Instream parameters are controlled by watershed parameters. (2) Wide ranges of watershed and instream factors have cumulative effects on sediment supply of streams.

Gravel-bed rivers at the reach scale

57

Physical models of river reaches have been used very successfully in New Zealand. Rob Ferguson did not mention them in his overview. However, we have a long way to go before numerical models will give us a good representation of the reality of river dynamics. There should still be a place for physical modelling. For coarse gravel rivers, adequate scaling is relatively easily achieved, and provided you have a large shed and pumps on hand, physical models are not that expensive.

Discussion by A. Papanicolaou The author should be complimented for addressing one of the most intriguing problems in modern sediment transport theory, viz. limitations of traditional regime theories and the need to look for an ‘outside the box’ approach by treating the bed characteristics within a reach scale as a degree of freedom characteristics. The author eloquently discusses the necessity for introducing 2-D dimensional models to describe the spatial variability within a reach scale triggered by the self organization of the bed, and the three way constant feedback process among ﬂow, banks, and bed morphology that is especially pronounced in rivers that undergo a system-wide adjustment. Another issue that the discusser, though, believes should be further discussed amongst the members of this community is the issue of sediment supply within a reach scale. Traditionally sediment supply is provided from upstream of a stream reach by assuming that is originated from instream sediment sources. Recently, lateral sediment contributions from the banks have been accounted for by introducing the 1.5-D models (Wu et al., 2004). However, sediment sources originated from hillslopes and ﬂoodplains (the uplands) have not yet been considered (e.g., Gouthard et al., same issue). Upland sediment delivery transport, in some cases, could account for up to 80% of the total lateral sediment inputs within a stream reach (McCool et al., 2000). In an effort to generate new knowledge and improved understanding of the complex interrelationships between watershed and instream parameters and the scale integrity inﬂuences on channel morphology, an integrated watershed hydrologic/ sedimentation framework for mountainous watersheds must be considered. This framework must utilize advanced analytical techniques and physically based numerical models for simulating upland (macro level) and instream (micro level) processes in an integrated fashion (Wang, 2005). As a ﬁrst attempt in this direction, Papanicolaou et al. (2003) have conducted a watershed ecosystem study in the South Fork of the Clearwater River in Idaho, USA. They identiﬁed the interdependencies between watershed and instream parameters and the clusters of parameters that primarily affect sediment supply and ultimately ﬁsh health. Papanicolaou et al. (2003) concluded the following (Fig. 2.8): (1) Instream parameters are controlled by watershed parameters. (2) Wide ranges of watershed and instream factors have cumulative effects on sediment supply of streams.

58

R. Ferguson

Figure 2.8. The ﬁgure illustrates the instream and watershed parameters that were found to affect sediment supply and aquatic health in the South Fork of the Clearwater River (SFCR).

(3) Lack of understanding of the complex interaction between watershed parameters with instream parameters and the effects of scale on ﬁsh species has caused, in some cases, the unsuccessful implementation and failure of these management plans (National Research Council, 1996). Based on cluster and factor statistical analysis performed over 50 parameters for about 100 years of record, it was determined that anthropogenic disturbances, watershed characteristics, river hydrology, and geometry have a signiﬁcant effect on sediment delivery. Speciﬁcally, for the SFCR it is shown that the governing parameters affecting sediment supply in that basin over a period of 50 years are the Mining and Debris processes, Fires and the erosion associated with them, Hillslope and other watershed physigraphical characteristics, River density, River discharge, River bank composition, and longitudinal slope. Reach scale models (e.g., 2-D or 3-D), therefore, need to account for those inputs through a coupling approach, for example, with watershed delivery models that are physically based (e.g., WEPP). Another shortcoming of the reach scale models is that they do not account for the presence of multi-modal type of distributions, especially when cohesive sediments are ubiquitous atop the river bed (Zeigler and Nisbet, 1994). In order for the reach scale to be able to truly model sand, gravel and cohesive sediment mixtures these models need to analyse either matrix-supported beds (clay and sand dominance over gravel) or clast-supported beds (gravel dominant).

References McCool, D.K., Pannkuk, C.D., Saxton, K.E., Kalita, P.K., 2000. Winter runoff and erosion on Northwestern USA cropland. Int. J. Sediment. Res. 15 (2), 149–161.

Gravel-bed rivers at the reach scale

59

National Research Council, 1996. Upstream Salmon Society in The Paciﬁc Northwest. National Academy Press, Washington, DC. Papanicolaou, A., Bdour, A., Evaggelopoulos, N., Tallebeydokhti, N., 2003. Watershed and stream corridor impacts on the ﬁsh population in the South Fork of the Clearwater River, Idaho. J. Am. Water Resour. Assoc. 39 (1), 191. Wang, S.S.Y., 2005. Integrated Modeling and Hydraulic Engineering. World Water Congress 173, 420. Wu, W., Dalmo, A.V., Want, S.S.Y., 2004. One-dimensional numerical model for nonuniform sediment transport under unsteady ﬂows in channel networks. J. Hydr. Eng. 130, 914. Zeigler, C.K., Nisbet, B.S., 1994. Fine-grained sediment transport in Pawtuxet River, Rhode Island. ASCE J. Hydraulic Eng. 120 (5), 561–576.

Reply by the author I thank the discussants for their interest and their contributions, all of which raise useful points which were omitted in my necessarily abbreviated survey of a very broad topic. Three of the discussions relate to the new data acquisition techniques which I highlighted in Section 3 of the paper. Milan et al. draw attention to something I did not mention in my outline of new approaches to quantifying morphology: the different combinations of time and space resolution with which this can be done. Their diagram shows this well, though I suspect it will be rendered out of date by future developments – just as my text was already out of date in not mentioning oblique laser scanning, which as they demonstrate extends the envelope of space–time resolution. An issue which is implicit in such developments, and which Lamarre and Roy explicitly raise, is to what extent we actually need the highest possible spatial resolution when seeking to understand reach-scale dynamics. This is an important question, and not one with a simple answer. To take their example of individual pebble clusters: it seems intuitively reasonable to neglect these as elements of the reach planform, but the effects of bed microtopography on ﬂow resistance do need to be taken into account somehow. Usually this is done by treating individual roughness elements as sub-grid-scale features whose combined effect is parameterised through a high value of Manning’s n or the Nikuradse roughness height. But some scientists working at the frontiers of ﬂow modelling would argue that a more rigorous approach is to model the effects of obstacle drag and wakes explicitly in a CFD model with very detailed bed geometry (e.g., Lane et al., 2004). My own feeling is that there will always be a place for parametric representations but that it would be proﬁtable to use reductionist process models to try to constrain and improve our parameterisations. Lamarre and Roy also point out that not all of the new techniques can be used in all kinds of rivers. As they say, dense vegetation and very coarse bed material pose particular problems, and traditional techniques remain valuable in such situations. I agree, and note the implication that Milan et al.’s scale diagram shows only the potential space–time range of each technique, not what is actually possible in a given situation. However, I don’t think this alters my general point that the available toolkit is much bigger than it was a decade ago, opening up new opportunities.

Gravel-bed rivers at the reach scale

59

National Research Council, 1996. Upstream Salmon Society in The Paciﬁc Northwest. National Academy Press, Washington, DC. Papanicolaou, A., Bdour, A., Evaggelopoulos, N., Tallebeydokhti, N., 2003. Watershed and stream corridor impacts on the ﬁsh population in the South Fork of the Clearwater River, Idaho. J. Am. Water Resour. Assoc. 39 (1), 191. Wang, S.S.Y., 2005. Integrated Modeling and Hydraulic Engineering. World Water Congress 173, 420. Wu, W., Dalmo, A.V., Want, S.S.Y., 2004. One-dimensional numerical model for nonuniform sediment transport under unsteady ﬂows in channel networks. J. Hydr. Eng. 130, 914. Zeigler, C.K., Nisbet, B.S., 1994. Fine-grained sediment transport in Pawtuxet River, Rhode Island. ASCE J. Hydraulic Eng. 120 (5), 561–576.

Reply by the author I thank the discussants for their interest and their contributions, all of which raise useful points which were omitted in my necessarily abbreviated survey of a very broad topic. Three of the discussions relate to the new data acquisition techniques which I highlighted in Section 3 of the paper. Milan et al. draw attention to something I did not mention in my outline of new approaches to quantifying morphology: the different combinations of time and space resolution with which this can be done. Their diagram shows this well, though I suspect it will be rendered out of date by future developments – just as my text was already out of date in not mentioning oblique laser scanning, which as they demonstrate extends the envelope of space–time resolution. An issue which is implicit in such developments, and which Lamarre and Roy explicitly raise, is to what extent we actually need the highest possible spatial resolution when seeking to understand reach-scale dynamics. This is an important question, and not one with a simple answer. To take their example of individual pebble clusters: it seems intuitively reasonable to neglect these as elements of the reach planform, but the effects of bed microtopography on ﬂow resistance do need to be taken into account somehow. Usually this is done by treating individual roughness elements as sub-grid-scale features whose combined effect is parameterised through a high value of Manning’s n or the Nikuradse roughness height. But some scientists working at the frontiers of ﬂow modelling would argue that a more rigorous approach is to model the effects of obstacle drag and wakes explicitly in a CFD model with very detailed bed geometry (e.g., Lane et al., 2004). My own feeling is that there will always be a place for parametric representations but that it would be proﬁtable to use reductionist process models to try to constrain and improve our parameterisations. Lamarre and Roy also point out that not all of the new techniques can be used in all kinds of rivers. As they say, dense vegetation and very coarse bed material pose particular problems, and traditional techniques remain valuable in such situations. I agree, and note the implication that Milan et al.’s scale diagram shows only the potential space–time range of each technique, not what is actually possible in a given situation. However, I don’t think this alters my general point that the available toolkit is much bigger than it was a decade ago, opening up new opportunities.

60

R. Ferguson

Rennie ampliﬁes the very brief mention in my text of his novel and highly promising method for estimating local bedload transport intensity and direction from acoustic Doppler ﬂow proﬁles. I agree with him that if the calibration problem can be overcome, this will give us a tremendous opportunity to constrain the least reliable part of any distributed morphodynamic model. The other two discussions relate to Section 4 on modelling. Williams asserts the ongoing value of mobile-bed physical modelling, especially in relation to observing what goes on in ﬂood conditions. I noted in Section 2.1 of my paper that laboratory experiments have made a major contribution to understanding planform self-organisation, and I did not mean to imply that the value of physical modelling stops there. As with ﬁeld and numerical approaches, though, it does have its limitations: it is hard to build up a picture of what is generic behaviour and what is speciﬁc to the particular conﬁguration studied; dynamical scaling is not completely unproblematic (e.g., the ﬁne tail of the prototype bed size distribution often has to be truncated); the initial bed is often looser than in nature so that transport rates are excessive (e.g., Monteith and Pender, 2005); and there is the issue of whether to feed or recirculate sediment, and if feed, at what rate and with what GSD. The question of sediment supply to a reach is the theme of Papanicolaou’s discussion, which expands on my very brief mention of this boundary condition which is often crucial in any model of bedload transport or morphodynamics. Papanicolaou draws attention to the possible signiﬁcance of lateral supply from hillslopes or ﬂoodplains, as well as the instream supply to the head of the reach, and sees a need for coupling channel reach models with basin- (watershed-) scale models of sediment delivery. This is clearly essential in trying to model the complete sediment budget. How important it is for understanding channels at reach scale depends on the calibre of the lateral supply, which if ﬁne becomes washload without much effect on the channel. Coarse lateral inputs are possible from steep tributaries, and also from bank erosion if the ﬂoodplain is mainly the product of lateral accretion and channel switching rather than vertical accretion. Indeed, in my ﬁeldwork experience much of the bedload in braided rivers is derived from bank retreat rather than bed scour.

References Lane, S.N., Hardy, R.J., Ingham, D.B., Elliott, L., 2004. Numerical modeling of ﬂow processes over gravelly surfaces using structured grids and a numerical porosity treatment. Water Resour. Res. 40 (1), art. no. W01302. Monteith, H., Pender, G., 2005. Flume investigations into the inﬂuence of shear stress history on a graded sediment bed. Water Resour. Res. 41 (12), art. no. W1240.

Gravel-Bed Rivers VI: From Process Understanding to River Restoration H. Habersack, H. Pie´gay, M. Rinaldi, Editors r 2008 Elsevier B.V. All rights reserved.

61

3 Hydrodynamics of gravel-bed rivers: scale issues Vladimir Nikora

Abstract The paper discusses several issues of gravel-bed river hydrodynamics where the scale of consideration is an inherent property. It focuses on two key interlinked topics: velocity spectra and hydrodynamic equations. The paper suggests that the currently used three-range spectral model for gravel-bed rivers can be further reﬁned by adding an additional range, leading to a model that consists of four ranges of scales with different spectral behaviour. This spectral model may help in setting up scales of consideration in numerical and physical simulations as well as in better deﬁning relevant ﬂuid motions associated with turbulence-related phenomena such as sediment transport and ﬂow–biota interactions. The model should be considered as a ﬁrst approximation that needs further experimental support. Another discussed topic relates to the spatial averaging concept in hydraulics of gravel-bed ﬂows that provides double-averaged (in time and in space) transport equations for ﬂuid momentum (and higher statistical moments), passive substances, and suspended sediments. The paper provides several examples showing how the double-averaging methodology can improve description of gravel-bed ﬂows. These include ﬂow types and ﬂow subdivision into speciﬁc layers, vertical distribution of the double-averaged velocity, and some consideration of ﬂuid stresses.

1.

Introduction

The key feature that makes river ﬂow different from other ﬂow types is the interaction between ﬂowing water and sedimentary bed. This interaction occurs over a wide range of scales, from the scale of a ﬁne sediment particle to the basin scale. A small-scale subrange of this wide range of scales is formed by turbulence and turbulence-related processes. This subrange extends from sub-millimetres to a channel width and covers motion of sediment particles in individual and collective (bedforms) modes, mixing and transport of various substances (e.g., nutrients, contaminants) and ﬂow–biota interactions. These turbulence-related phenomena E-mail address: [email protected] ISSN: 0928-2025

DOI: 10.1016/S0928-2025(07)11113-5

62

V. Nikora

are especially important in the functioning of gravel-bed rivers and therefore constantly attract researchers’ attention, consistently forming a topic of discussion at Gravel-Bed Rivers Workshops (e.g., Livesey et al., 1998; Nelson et al., 2001; Roy and Bufﬁn-Belanger, 2001; Wilcock, 2001). At present, turbulence research of gravel-bed rivers is based on two fundamental physical concepts: eddy/energy cascade and coherent structures. Originally these concepts have been developed independently, and it is only recently that researchers started viewing them as interlinked phenomena. These concepts, together with fundamental conservation equations for momentum, energy, and substances, represent two facets of ﬂow dynamics and two corresponding research approaches: statistical and deterministic. The deterministic approach stems from some ‘coherency’ in turbulent motions and from hydrodynamic equations based on fundamental conservation principles, while the statistical approach recognises ‘irregular’ components in hydrodynamic ﬁelds and therefore focuses on their statistical properties. The statistical approach is based on two important procedures: (1) decomposition of hydrodynamic ﬁelds into slow (or mean) and fast (or turbulent) components; and (2) averaging or ﬁltering of instantaneous variables and corresponding hydrodynamic equations. The ﬁrst procedure is known as the Reynolds decomposition in the case of time and ensemble (i.e., probabilistic) averaging and as Gray’s (1975) decomposition in the case of spatial averaging. This procedure can be interpreted as a scale decomposition or separation of scales. The second procedure can be formulated in many different ways among which time, ensemble, and area/volume averaging are most common. This second procedure can be viewed as a scaling-up procedure that changes the scale of consideration from one level in time–space-probability domain to another level. In this respect, scale is an inherent feature of any hydrodynamic equation, which is not always recognised in Earth Sciences. The generalised hydrodynamic equations formulated in terms of statistical moments of various orders were ﬁrst proposed by A.A. Friedman and L.V. Keller in the 1920s (Monin and Yaglom, 1971). As an example, the well-known Reynolds averaged Navier–Stokes equation represents an equation for the ﬁrst-order moments of velocity and pressure ﬁelds. Another direction within the statistical approach is formulation of statistical turbulence theories based on physical intuition rather than on basic conservation principles expressed by hydrodynamic equations. A well known example is Kolmogorov’s turbulence theory and its associated ‘‘5/3’’ law for the inertial subrange where energy is transferred from larger scales to smaller scales without dissipation and/or additional production. Scale is an inherent feature in such theories as well. It can be argued that the currently popular terms in Earth Sciences such as scaling, scale-invariance, self-similarity, characteristic scales, and scaling behaviour largely stem from these statistical theories of turbulence (e.g., Barenblatt, 1995, 2003). The range of problems and concepts related to gravel-bed river turbulence is wide and it is impossible to address them all within a single paper. Instead, following the central theme of this Workshop, the paper will review topics where the scale issue, as described above, is a fundamental feature, proper account of which may improve current understanding of gravel-bed rivers dynamics. The velocity spectra in gravelbed rivers will be discussed ﬁrst as it forms a general framework for multi-scale considerations. With this as background, a brief discussion on how time and spatial

Hydrodynamics of gravel-bed rivers

63

scales are associated with currently used hydrodynamic equations will follow. This will lead to a more detailed consideration of the double-averaging methodology dealing with hydrodynamic equations averaged in both time and space. In the author’s own research, this methodology evolved in the mid-1990s when he tried to use conventional Reynolds averaged equations to study near-bed region of gravel-bed ﬂows and found them inconvenient because of scale inconsistency. The paper will conclude with several examples showing how the double-averaging methodology can improve description of gravel-bed ﬂows. The examples include ﬂow types and ﬂow subdivision into speciﬁc layers, vertical distribution of the double-averaged velocity, and some consideration of ﬂuid stresses. The examples support a view that this methodology opens a new perspective in gravel-bed rivers research and may help in clarifying some long-standing problems. There are many other important aspects of gravel-bed river turbulence that are not covered in this paper. Interested readers will beneﬁt from checking a comprehensive report of Lopez and Garcia (1996) and very recent reviews of the problem given in Roy et al. (2004, and references therein) and Lamarre and Roy (2005, and references therein).

2.

Velocity spectra in gravel-bed rivers

Velocity ﬂuctuations in gravel-bed rivers cover wide ranges of temporal and spatial scales, from milliseconds to many years and from sub-millimetres to tens of kilometres. The smallest temporal and spatial scales relate to the so-called dissipative eddies through which energy dissipation occurs due to viscosity. The largest temporal scales of velocity ﬂuctuations relate to long-term (climatic) ﬂuctuations of the ﬂow rate, while the largest spatial ﬂuctuations are forced by morphological features such as meanders or even larger structures of, for example, tectonic origin. The amplitude of velocity ﬂuctuations typically increases with period and wavelength (i.e., with the scale). This dependence can be conveniently summarised using velocity spectra showing how the energy of ﬂuctuations is distributed across the scales (Fig. 3.1). The spectra in Fig. 3.1 represent a result of conceptualisation of extensive turbulence and hydrometric measurements (Grinvald and Nikora, 1988). The low frequency (large periods) range in the frequency spectrum is formed by intra-annual and inter-annual hydrological variability while high-frequency (small periods) range is formed by ﬂow turbulence (Fig. 3.1a). The connection between these two extreme ranges is not yet clear and may relate to various large-scale ﬂow instabilities (Grinvald and Nikora, 1988), deﬁned in Fig. 3.1a as ‘‘hydraulic phenomena’’. The low wave-number (large spatial scale) range in the wave-number spectrum is formed by morphological variability along the ﬂow such as bars and/or meanders (Fig. 3.1b), as was mentioned above. At small spatial scales (less than ﬂow width) velocity ﬂuctuations are due to turbulence. If the wave-number and frequency turbulence spectra can often be linked through Taylor’s ‘frozen’ turbulence hypothesis (as can be seen in Figs. 3.1a and b, Nikora and Goring, 2000a), the relationship between large-scale ranges of the wave number and frequency spectra are not as clear. The turbulence ranges in Figs. 3.1a and b can be conceptually subdivided into macro-turbulence (between ﬂow depth

V. Nikora

Weeks

Geomorphological variability

W/U H/U Time scale

Z/U

∆/U (ν/ε)0.5

b Micro

Meso

Intermediate subrange

W H Spat ial scale

Z

Dissipative range

Inertial subranges

Inertial subranges Large-eddies range

Large forms (e.g., meanders)

Intra-channel forms (e.g., bars) ~10W

Micro

Turbulence Macro

~10Wo

a

Meso

Intermediate subrange

?

Macro

Large-eddies range

Intra-annual variability

Hydraulic phenomena Years

Velocity Frequency Spectrum

Turbulence

∆

Dissipative range

Hydrological variability

Inter-year variability

Velocity Frequency Spectrum

64

(ν3/ε)1/4

Figure 3.1. Schematised velocity spectra in gravel-bed rivers: (a) frequency spectrum; and (b) wavenumber spectrum (W0 and W are the river valley and river channel widths, respectively).

and ﬂow width), meso turbulence (between dissipative scale and ﬂow depth), and micro turbulence (dissipative eddies). There may be a variety of energy sources for ﬂow turbulence with the key source being the energy of the mean ﬂow, which is transferred into turbulent energy through velocity shear and through ﬂow separation behind multi-scale roughness elements. In the velocity spectra, the ﬁrst transfer occurs at the scale of the ﬂow depth while the second transfer occurs at the scale of roughness size(s) D (Figs. 3.1a and b). The importance of a particular range in turbulence dynamics and speciﬁc boundaries of spectral ranges should depend on width to depth ratio and relative submergence. The information on turbulence spectra in gravel-bed rivers is very fragmentary and mainly covers the longitudinal velocity component u in the range of scales from approximately one tenth of depth to several depths (e.g., Grinvald and Nikora, 1988; Nezu and Nakagawa, 1993; Roy et al., 2004). It has been shown that this region usually covers the inertial subrange where velocity spectra follow Kolmogorov’s ‘‘5/3’’ law. In most such studies a three-range model of spectra has been accepted, implicitly or explicitly, which consists of: (1) the production range where spectral behaviour has not been identiﬁed speciﬁcally; (2) the inertial subrange where spectra

Hydrodynamics of gravel-bed rivers

65

follow the ‘‘5/3’’ law (there is no energy production or dissipation in this subrange; Monin and Yaglom, 1975); and (3) the viscous range where spectral density decays much faster than in the inertial subrange. This conceptual model stems from Kolmogorov’s concept of developed turbulence (i.e., at sufﬁciently large Reynolds number; Monin and Yaglom, 1975). However, the true spectral behaviour outside the range of length scales from E0.1 to E(2 to 3) ﬂow depths, although very important for engineering and ecological applications, is not yet clear. Below, this range of scales is revised and extended using physical and scaling arguments, and then compared with available measurements. The analysis begins with the reasonable assumption that velocity spectra Sij (k) in high Reynolds number gravel-bed ﬂows with dynamically completely rough beds are fully determined by one velocity scale (i.e., the shear velocity u), and three characteristic length scales: (1) characteristic bed particle size (or roughness length) D, assuming that it essentially captures the effects of bed particle size distribution; (2) distance from the bed z (see Section 4.3 for a discussion of bed origin); and (3) mean ﬂow depth H. These are the main scales for ﬂows over both ﬁxed and mobile beds. The bed conditions for our considerations are somewhat simpliﬁed, i.e., channel width and characteristic scales of bed-forms are excluded from our analysis. Also, the viscous range of scales is not considered. This range, although important for dissipation mechanisms (which are beyond the scope of this paper), contributes very little to the total spectral energy. With these assumptions one can have: Sij ðkÞ ¼ F ðu ; D; z; H; kÞ

(3.1)

where k is longitudinal wave number in the direction of the mean ﬂow (k ¼ 2p/l, l is an eddy characteristic scale in the streamwise direction). After applying conventional dimensional analysis relationship (3.1) reduces to: Sij ðkÞ ¼ u2 k1 f ðkD; kz; kHÞ

(3.2)

Using (3.2) one may consider spectral behaviour of (i) the largest eddies (l4a1iH), (ii) intermediate eddies (a2i zoloa1i H), and (iii) relatively small eddies (a3i Doloa2i z) where a1i, a2i, and a3i are scaling coefﬁcients for the ith velocity component (i ¼ 1 for the longitudinal component u, i ¼ 2 for the transverse component v, and i ¼ 3 for the vertical component w). For the largest eddies (l4a1i H4a2i z4a3i D), i.e.: H H z z kHo2pb1i o2pb2i o2pb3i or kzo2pb1i o2pb2i o2pb3i ; z D H D (3.3) 1 where bki ¼ aki incomplete self-similarity in kH (or self-similarity of the second kind after Barenblatt, 1995, 2003), and complete self-similarity in kD and kz (note that kDokzokH) are assumed. The latter means that at small kD and kz the spectrum does not depend on these variables and they can be dropped while the former means that at a small ratio H=l / kH we may present S ij ðkÞ ¼ f ðkHÞ as Sij ðkÞ / ðkHÞa . In the case of the complete similarity in kz the contributions to spectra from the largest eddies are invariant with respect to distance from the bed, which seems physically reasonable (e.g., Kirkbride and Ferguson, 1995; Nikora and Goring, 2000b; Roy et al., 2004;

V. Nikora

66 Nikora, 2005). All these reduce (3.2) to: Sij ðkÞ ¼ c1ij u2 k1 ðkHÞa

(3.4)

where c1ij is a constant. Relationship (3.4) can be further simpliﬁed using a physical argument that the largest eddies represent a link between the mean ﬂow and turbulence, i.e., in spectra they occupy the region of turbulence energy production where eddies interact with the mean ﬂow and between themselves. This energy exchange between large eddies suggests that their spectral contributions are invariant with wave number and so k should be dropped from (3.4). This assumption gives a ¼ 1 and simpliﬁes relationship (3.4) to a form: z 1 (3.5) Sij ðkÞ ¼ c1ij u2 H or Sij ðkHÞ ¼ c1ij u2 or S ij ðkzÞ ¼ c1ij u2 H which is valid for kzo2pb1i ðz=HÞ (see (3.3)). For the intermediate eddies (a2i zoloa1i H), i.e.: z z 2pb1i okzo2pb2i o2pb3i (3.6) H D complete self-similarity with kD, kz, and kH is assumed that reduces (3.2) to the relationship: Sij ðkÞ ¼ c2ij u2 k1

or

S ij ðkHÞ ¼ c2ij u2 ðkHÞ1

or S ij ðkzÞ ¼ c2ij u2 ðkzÞ1

(3.7)

which is valid for 2pb1i ðz=HÞokzo2pb2i and where c2ij is a constant. Relationships (3.6) and (3.7) mean that eddies from this range of scales are independent of the characteristic scales D, z, and H and depend only on the velocity scale, i.e., u . Finally, for relatively small eddies (a3i Doloa2i z), i.e.: z z 2pb1i o2pb2i okzo2pb3i (3.8) H D incomplete self-similarity with kz and complete self-similarity with kD and kH are assumed, i.e.: Sij ðkÞ ¼ c3ij u2 k1 ðkzÞb

(3.9)

where c3ij is a constant. To deﬁne an exponent b one may use a reasonable assumption that these eddies form the inertial subrange, i.e., S ii ðkÞ / k5=3 and C uw ðkÞ / k7=3 (Monin and Yaglom, 1975) where Sii ðkÞ is the auto-spectrum for the ith velocity component (i.e, the spectrum of a single velocity component), and C uw ðkÞ is the co-spectrum, which is the real part of the cross-spectrum of longitudinal and vertical velocities. The analysis here is restricted to just this one off-diagonal component of the spectral tensor since C uw ðkÞ provides important information Ron contributions 1 from different eddies to the primary shear stress u0 w0 ; i:e:; u0 w0 ¼ o C uw ðkÞdk. This assumption gives b ¼ 2=3 for the auto-spectra and b ¼ 4=3 for the co-spectra and simpliﬁes (3.9) to the following relationships: Sii ðkÞ ¼ c3ii u2 k5=3 z2=3 C uw ðkÞ ¼ c3uw u2 k7=3 z4=3

S ii ðkzÞ ¼ c3ii u2 ðkzÞ5=3

or or

C uw ðkzÞ ¼ c3uw u2 ðkzÞ7=3

(3.10) (3.11)

Hydrodynamics of gravel-bed rivers

67

which are valid for 2pb2i okzo2pb3i z=D. Note that the distance z from the bed in (3.8) may be interpreted as an ‘external’ Kolmogorov’s scale (Monin and Yaglom, 1975) deﬁned by the size of ‘attached’ eddies (i.e., eddies ‘growing’ from the bed; the attached eddies hypothesis was ﬁrst introduced by Townsend, 1976). This suggests that k23 z ¼ const where k23 is the low-wave-number limit for the inertial subrange, i.e., the boundary between (3.7) and (3.10). The performance of the above relationships for individual velocity components, wave-number limits for (3.5), (3.7), (3.10), and (3.11) and ‘universal’ constants c1ij , c2ij , and c3ij should be deﬁned from experiments. The above conceptual model consists of four ranges of scales with different spectral behaviour: (I) the range of the largest eddies (l4a1i H) with Sij ðk; zÞ / k0 z0 ; (II) the range of intermediate eddies (a2i zoloa1i H) with Sij ðk; zÞ / k1 z0 ; (III) the range of relatively small eddies (a3i Doloa2i z) with S ii ðk; zÞ / k5=3 z2=3 and C uw ðk; zÞ / k7=3 z4=3 (known as the inertial subrange where no energy production or dissipation occurs); and (IV) the viscous range (not speciﬁed here). In addition to previous three-range concepts for open-channel ﬂows (e.g., Grinvald and Nikora, 1988; Nezu and Nakagawa, 1993) this model speciﬁes the spectral behaviour at very low wave numbers and adds an additional spectral range with S ii ðkÞ / k1 (Fig. 3.2a). If spectral ranges (I), (III), and (IV) are well known and are widely used in physical considerations, range (II) with Sij ðkÞ k1 is much less known in gravel-bed rivers research. Its physical origin is still unclear (see, e.g., Yaglom, 1993; Katul and Chu, 1998 for various concepts and associated references). A plausible physical mechanism that may explain the appearance of this spectral range is brieﬂy reviewed below, following Nikora (1999). There are two important properties of wall turbulence (close to the bed, within the logarithmic layer which is assumed to exist) which are well tested and accepted in wall turbulence studies. A. The shear stress t is approximately constant and equal to t ¼ ru2 (u is the friction velocity, and r is ﬂuid density). The production of the total turbulence energy P is approximately equal to the energy dissipation d that leads to P d u3 =z. These properties describe Townsend’s (1976) equilibrium wall layer with constant shear stress. B. The mean ﬂow instability and velocity shear generate a hierarchy of eddies attached (in the sense of Townsend, 1976) to the bed so that their characteristic scales are proportional to the distance z from the bed. Using property B it can reasonably be assumed that, due to ﬂow instability and velocity shear, the energy injection from the mean ﬂow into turbulence occurs at each distance z from the wall, with generation of eddies with characteristic scale L z. These eddies transfer their energy at rate e to smaller eddies and may be viewed as energy cascade initiators. In other words, it is suggested here that at each z a separate Kolmogorov’s cascade is initiated which is superposed with other energy cascades initiated at other z’s. As a result of this superposition of cascades, the energy dissipation ed at a particular distance z presents a superposition of down-scale energy ﬂuxes, e, generated at this and at larger z (contribution from cascades generated at smaller z is negligible; justiﬁcation for this may be found in Townsend, 1976). Thus,

V. Nikora

68 a

ε(kz)~(u*3/ z )(kz) ε(kz) = εd ~ u*3/ z

S(kz)~(kz)-1

S(kz)~(kz)-5/3

3

2

1

Ln [ε ( k z )]

Ln [Sii( k z)]

ε(kz)

4 S (kz)

kz~1

kz~(z/H)

Ln [k z]

b

100

S ~k-1

S ~ k-5/3

10 Sij ( kz ) /u*2 =u Sij ( 2πf ) / z u*2

1

0.0095 0.0380 0.4670

0.01 0.01 S

0.1

~k-1

1 S ~ k-5/3

10

S ~ k-1

10

S ~k-5/3

v

1

z/H

0.1

100

100

u

0.1 0.01 0.01

10 10

w

1 S ~k-7/3

1

1

0.1

0.1

0.01 0.001

0.01

0.1

S ~k-1

10

uw

0.0001 0.01

0.1

1

kz = 2π f z / u

10

0.01

0.1

1

10

kz = 2π f z / u

Figure 3.2. (a) Schematised velocity auto-spectra Sii ðkzÞ and energy transfer rate ðkzÞ showing: (1) the large-scale energy production range (k4H 1 ); (2) the ‘‘1’’ scaling range (H 1 okoz1 ) where energy cascades initiated at each z are superimposed and ðkÞ changes as ðkÞ k; (3) the inertial subrange (k4z1 ) which results from superposition of inertial subranges generated at each z and, therefore, ðkÞ d ; and (4) the dissipative range. (b) An example of velocity spectra at z/H ¼ 0.0095, 0.0380, and 0.4670, measurements were made with acoustic Doppler velocimeters (ADV) with the sampling frequency of 25 Hz and duration of 20 min, 1997, Balmoral Canal (ﬂow rate ¼ 5.14 m3/s; cross-sectional mean velocity ¼ 1.05 m/s; cross-sectional mean depth ¼ 0.78 m), New Zealand [see also Nikora (2005) for more details].

the energy ﬂux e across the scales at any z depends on the scale under consideration, i.e., on wave number k. The ﬂux e increases with k until it reaches 2p/z and then, for k ð2p=zÞ, stabilises being equal to ed (Fig. 3.2a). In other words, at a given distance zg the energy ﬂux ðkÞ for ko2pz1 g represents the energy dissipation ed observed at

Hydrodynamics of gravel-bed rivers

69

z ¼ 2pk1 , z4zg . Using property A (i.e., d u3 =z) and bearing in mind that L z k1 , we have ðkÞ u3 k for ð2p=HÞ k ð2p=zg Þ. The scale H is an external scale of the ﬂow. Following this phenomenological concept and using the 2=3 4=3 7=3 one can inertial subrange relationships S ii ðkÞ d k5=3 and C uw ðkÞ d u2 k obtain (3.7), (3.10) and (3.11). Thus, the existence of the ‘‘1’’ spectral law in wallbounded turbulence is explained by the effect of superposition of Kolmogorov’s energy cascades generated at all possible distances from the wall, within an equilibrium layer. This concept is justiﬁed using only the well-known properties of wallbounded ﬂows. The energy cascades initiated at any z may be linked to large eddies attached to the bed and scaled with z. Such eddies may be associated with coherent structures, considered for example in Roy et al. (2004). Indeed, the data presented in Nikora (2005) suggest that the clusters of bursting events are the main contributors to range I with Sij ðkÞ / u2 H while range II with S ij ðkÞ / u2 k1 is probably formed by individual bursting events. The latter may be viewed as the energy cascade initiators. The four-range model described above has been well supported by data from gravel-bed ﬂows (e.g., Nikora and Smart, 1997; Nikora and Goring 2000b; Roy et al., 2004). As an illustration, Fig. 3.2b shows normalised spectra Sij ðkzÞ=u2 for three representative values of z (so that effects of normalisation can be clearly seen without attenuation by numerous curves) measured with Acoustic Doppler Velocimeters in a gravel-bed Balmoral Canal (New Zealand). It is evident from Fig. 3.2b that the measured spectra do support (3.7) and (3.10) for all three velocity components, although the ‘‘1’’ ranges for the transverse and vertical velocities are fairly narrow. Besides, this ﬁgure also supports scaling relationships (3.7) and (3.11) for the co-spectra. At low wave numbers all spectra tend to constant values as predicted by (3.5). The typical values for the constants c1ij, c2ij, and c3ij obtained for gravel-bed ﬂows are c1uu 1:0, c1vv ¼ 0:13, c1ww ¼ 0:04 [see equation (3.5)], c2uu ¼ 0:90, c2vv ¼ 0:50, c2ww ¼ 0:30 [see equation (3.7)], and c3uu ¼ 0:90, c3vv ¼ 1:20, c3ww 0:9 [see equation (3.10)]. The standard measurement errors of the above values are within 5–25%. Note that the ratio c3ww =c3uu does not satisfy Kolmogorov’s theory of locally isotropic turbulence, i.e., c3ww =c3uu 1:0o4=3, which is probably due to deviation from local isotropy. The satisfactory agreement between the proposed scaling model and measurements for the intermediate ﬂow region (not very close to either the bed or the water surface) may have immediate applications for the broad-band turbulence intensities. Indeed, the integration of the total spectra for velocity components u, v, and w gives: 2 z z si K ¼ M i N i ln ¼ 1:84 1:02 ln and (3.12) u H u2 H where si is the standard deviation of ith velocity component; K ¼ 0:5ðs2u þ s2v þ s2w Þ is the total turbulent energy; and M u ¼ 1:90, M v ¼ 1:19, M w ¼ 0:59, N u ¼ 1:32, N v ¼ 0:49, and N w ¼ 0:22 are derived from ﬁeld experiments (Nikora and Goring, 2000b). Equations (3.12) show how the turbulence intensity and energy change with changing distance from the bed. Although the above conceptual model for velocity spectra in gravel-bed ﬂows is plausible and well supported by data for particular hydraulic conditions, it should be treated as a preliminary result rather than a solution of the problem. Indeed, the

70

V. Nikora

model, while applicable for some conditions, has many limitations and does not cover many other possible scenarios encountered in the ﬁeld. The future improvements should account for the effects of relative submergence, width to depth ratio, multi-scale bed forms, wake turbulence, aquatic vegetation, bed permeability, topology of coherent structures, and other factors.

3.

Scales, hydrodynamic equations, and the double-averaging methodology

Velocity ﬂuctuations in gravel-bed rivers, highlighted in the previous sections, form a wide continuous spectrum that makes statistical approach for their description and prediction inevitable. Although ideally it would be preferable to study the whole range of scales simultaneously (i.e., resolving the smallest temporal and spatial scales involved), in practical terms it is impossible and often is not necessary. In principle, the small-scale effects can be incorporated into larger-scale dynamics by integrating corresponding hydrodynamic equations. This procedure, as already mentioned, is commonly formulated as either time or ensemble or area/volume averaging, or combination of them. This scaling-up procedure is inbuilt into currently used hydrodynamic equations. Indeed, depending on temporal and spatial resolution these equations can be broadly classiﬁed as: (1) equations with no time/ensemble and spatial averaging for (instantaneous) hydrodynamic variables (e.g., Navier–Stokes equation for momentum, NS); (2) spatially ﬁltered hydrodynamic equations for variables with small-scale spatial averaging (e.g., Large Eddy Simulation, LES; no time averaging is involved); and (3) time-(or ensemble) averaged hydrodynamic equations with no spatial averaging, known as the Reynolds Averaged NS equations (RANS). The spatial (LES) or time (RANS) averaging of NS equations for instantaneous variables can be viewed as a scaling-up procedure that changes the scale of consideration from a point in time-space (as in NS) to a larger spatial (as in LES) or temporal (as in RANS) scales. This classiﬁcation can be further extended by adding hydrodynamic equations for variables averaged in both time and space, which can be deﬁned as the double-averaged hydrodynamic equations (DANS). The double averaging upscales the original NS in both time and space domains. The selection of the equations for hydraulic modelling is often based, implicitly or explicitly, on scale considerations, i.e., bearing in mind velocity spectra considered in the previous section. To take all advantages provided by direct numerical solution (DNS) of the Navier–Stokes equations or LES one needs access to high-performance computing facilities as well as highly resolved initial and boundary conditions, which are unlikely to be available for many real-life engineering or ecological applications. Therefore, the RANS-based modelling approach is currently the most popular in solving practical problems although methodologically it is inconsistent in accounting for drag forces acting on the rough bed. In many real-life situations the commonly used RANS equations are difﬁcult to implement due to the highly three-dimensional small-scale structure of the mean ﬂow and turbulence, especially in the near-bed region (e.g., Lamarre and Roy, 2005). In addition, most applications deal with spatially averaged roughness parameters that cannot be linked explicitly with local

Hydrodynamics of gravel-bed rivers

71

(point) ﬂow properties provided by the Reynolds equations. A more straightforward approach is to use the DANS-based models, which are rigorously derived for roughbed ﬂows and provide explicit guidance in closure development and parameterisations. DANS-based models may successfully ﬁll a gap in modelling capabilities for gravel-bed ﬂow problems where the RANS-based models are not suitable. In the next paragraph we introduce the double-averaging methodology, which potentially may advance current understanding of gravel-bed ﬂow hydrodynamics as well as provide a practical modelling tool. The double-averaged equations for turbulent rough-bed ﬂows have been ﬁrst introduced and advanced by atmospheric scientists dealing with air ﬂows within and above terrestrial canopies such as forests or bushes (Wilson and Shaw, 1977; Raupach and Shaw, 1982; Finnigan, 1985, 2000). Later the double-averaging approach has been adopted in environmental hydraulics (e.g., Gimenez-Curto and Corniero Lera, 1996; McLean et al., 1999; Lopez and Garcia, 2001; Nikora et al., 2001, 2004, 2007a, b) but its applications for modelling, experimental design, and data interpretation are still largely undeveloped. This section will brieﬂy discuss the double-averaged momentum equation that then will be used to illustrate the advantages of this methodology. Similar equations can be also derived for conservation of mass, energy, and other velocity moments (e.g., turbulent shear stresses). In Nikora et al. (2007a) it has been shown that for a reasonably general case of static and mobile bed surfaces with roughness elements such as moving gravel particles the double-averaged (in time ﬁrst and in space second) momentum equation can be written as: @h¯ui i @h¯ui i 1 @fh¯pi 1 @fhu~ i u~ j i 1 @fhu0i u0j i ¼ gi þ h¯uj i @t @xj rf @xi f @xj f @xj 1 @ @ui þ f n @xj f @xj ZZ ZZ s s 11 1 1 1 @ui þ pni dS n nj dS @xj r f V0 f V0 S int

ð3:13Þ

S int

where ui is the ith component of the velocity vector; p is pressure; gi is the i-th component of the gravity acceleration; r is ﬂuid density; V0 is the total volume of the averaging domain (thin slab parallel to the mean bed); n is the inwardly directed unit vector normal to the bed surface (into the ﬂuid); Sint is the extent of water–bed interface bounded by the averaging domain; and f is the rough bed ‘porosity’ also deﬁned in Nikora et al. (2001) as the roughness geometry function; it is discussed in the next paragraph. The overbar and angle brackets in equation (3.13) denote time and spatial (volume) averaging, respectively. The superscript ‘‘s’’ denotes superﬁcial time average (Nikora et al., 2007a) when averaging time interval includes both periods when the spatial points are intermittently occupied by ﬂuid and when they are occupied by roughness elements (e.g., by moving gravel particles). Equation (3.13) uses Reynolds’ decomposition y ¼ y¯ þ y0 for instantaneous variables and an analogue of Gray’s (1975) decomposition for the time-averaged variables, ~ where y is a hydrodynamic variable. The wavy overbar denotes the ¯ þ y, y¯ ¼ hyi

72

V. Nikora

spatial ﬂuctuation in the time-averaged ﬂow variable, i.e., the difference between ~ ¼ 0), similar to ¯ and time-averaged y¯ values (y~ ¼ y¯ hyi, ¯ hyi the double-averaged hyi 0 ¯0 ¯ the conventional Reynolds decomposition of y ¼ y þ y ; y ¼ 0. The spatial averaging is often performed over a volume V0 that is a thin slab parallel to the mean (or ‘smoothed’) bed. The plane dimensions of the averaging domain should be larger than typical mean ﬂow heterogeneities, introduced by roughness, but much smaller than the large-scale features in bed topography. For gravel-bed rivers they should be much larger than gravel particles, but much smaller than sizes of rifﬂes or pools. Equation (3.13) has been derived in a single-step procedure from the Navier–Stokes equation for instantaneous variables using the averaging theorems linking double-averaged derivatives with derivatives of the double-averaged variables (Nikora et al., 2007a). It accounts for roughness mobility and change in roughness density with spatial coordinates and with time, which make them different from similar equations considered in terrestrial canopy aerodynamics and porous media hydrodynamics. The derivation of (3.13) accounted for both spatial porosity fs ¼ V f =V 0 and ‘time’ porosity ft ¼ T f =T, where Vf is the volume occupied by ﬂuid within an averaging (total) volume V0; T is the total averaging time interval including periods when the spatial points are intermittently occupied by ﬂuid and roughness elements (e.g., by moving gravel particles); and Tf is the averaging time interval equal to sum of time periods when a spatial point under consideration is occupied by ﬂuid only. In equation (3.13) it is assumed that ft does not (spatially) correlate with the ¯ ¼ hft ihyi). ¯ For many applications this time-averaged ﬂow parameters (i.e., hft yi assumption is reasonable and allows replacing the product fs hft i with a single symbol f ¼ fs hft i. For ﬁxed (static) roughness elements we have ft ¼ T f =T 1 and thus f ¼ fs . For gravel-bed ﬂows, the function fðzÞ changes upwards from the bed material porosity fs min deeply in the sediment layer to 1 at the roughness tops to zero at the air–water interface. In the case of a ﬂat water surface, there is a discontinuity in fðzÞ when it changes from 1 to 0. In the case of a disturbed water surface (e.g., random surface waves), the change in fðzÞ from 1 to 0 is likely to be smooth, similar to the water–sediment interface. Note that in previous work (e.g., Nikora et al., 2001), the roughness geometry function was deﬁned for area averaging and denoted by a symbol A. Here we use the symbol f to make distinction between area and volume averaging. In comparison with the conventional Reynolds-averaged Navier–Stokes equation, the proposed double-averaged momentum equation contains several additional terms which explicitly present dispersive or form-induced stresses ﬁelds, the form drag per unit hu~ i u~ j i due to spatial variations RR in time-averaged s ﬂuid volume f pi ¼ 1=ðfV 0 Þ Sint pni dS , and viscous drag per unit ﬂuid volume RR s f ni ¼ 1=ðfV 0 Þ Sint ðrn@ui =@xj Þnj dS . The quantities hu~ i u~ j i in equation (3.13) stem from spatial averaging, similar to u0i u0j in the Reynolds-Averaged equations which represent a result of time (ensemble) averaging of the Navier–Stokes equation for instantaneous variables. In other words, the double-averaged equations relate to the time-averaged equations in a similar way as the time-averaged equations relate to the equations for instantaneous hydrodynamic variables. Assessment of the signiﬁcance of these terms in equation (3.13) for different hydrodynamic and bed roughness

Hydrodynamics of gravel-bed rivers

73

conditions is currently underway (e.g., Nikora et al., 2007b). An important additional advantage of using double-averaged hydrodynamic parameters and equations is a better coupling between the surface water ﬂow and the sub-surface ﬂow within the porous bed where volume-averaged variables are traditionally used (e.g., Whitaker, 1999). The next section provides a brief review of several issues of gravel-bed ﬂow dynamics, which are discussed based on the double-averaging methodology and which illustrate its advantages.

4. 4.1.

Hydrodynamics of gravel-bed ﬂows: double-averaging perspective Vertical structure of gravel bed flows

Based on an analysis of the double-averaged momentum equation (3.13), Nikora et al. (2001, 2007b) suggested four types of rough-bed ﬂows (Fig. 3.3), depending on ﬂow submergence H m =D (Hm is the maximum ﬂow depth, i.e., the distance between water surface and roughness troughs). Here this classiﬁcation is adopted, with some modiﬁcations, for gravel-bed ﬂows. The ﬂow type I is the ﬂow with high relative submergence, which contains several layers and sublayers (neglecting viscous sublayers associated with gravel particles): (1) near-water-surface layer where ﬂow structure is inﬂuenced by the free surface effects such as turbulence damping and various types of water surface instabilities, and which for a dynamic non-ﬂat air–water interface may be further subdivided into an upper sub-layer with a smooth transition in fðzÞ from 1 (water) to 0 (air) where drag terms and hu~ i u~ j i in equation (3.13) may be important, and a lower sublayer where form-induced stresses hu~ i u~ j i may be essential (these sublayers are similar to the interfacial and form-induced sublayers at water–sediment interface described below); (2) outer or intermediate layer, where viscous effects and form-induced momentum ﬂuxes due to water surface disturbances and bed roughness are negligible, and the spatially averaged equations are identical to the time-averaged equations; (3) the z

Mean water surface

zws Near-water-surface layer

zint

Outer (or intermediate)layer

zL

Logarithmic layer zR zc zt

Roughness Form-induced sublayer Interfacial sublayer layer Subsurface layer

dφ =0 zf dz φ(z) 0 φmin 1

Flow Type I

Flow Type IIIFlow Flow Type IV Type II

Figure 3.3. Flow types and ﬂow subdivision into speciﬁc regions in gravel-bed ﬂows.

V. Nikora

74

logarithmic layer (as the relative submergence is large enough to form an overlap region) that differs from the outer layer by characteristic velocity and length scales; (4) the form-induced (or dispersive) sublayer, below the logarithmic layer and just above the roughness crests, where the time-averaged ﬂow may be inﬂuenced by individual roughness elements and thus the terms hu~ i u~ j i may become non-zero; (5) the interfacial sublayer, which occupies the ﬂow region between roughness crests and troughs and where momentum sink due to skin friction and form drag occurs; and (6) subsurface layer below the interfacial sublayer. The interfacial and form-induced sublayers, combined together, can be deﬁned as the roughness layer. The nearwater-surface layer can be viewed as a near-surface counterpart of the roughness layer. The other three ﬂow types are: (II) ﬂow with intermediate relative submergence consisting of the subsurface layer, a roughness layer, an upper ﬂow region which does not manifest a genuine universal logarithmic velocity proﬁle as the ratio H m =D is not large enough, and the near-water-surface layer; (III) ﬂow with small relative submergence with a roughness layer overlapped with the near-water-surface layer; and (IV) ﬂow over a partially inundated rough bed consisting of the interfacial sublayer overlapped with the near-water-surface layer (Fig. 3.3). These four ﬂow types and their subdivision into speciﬁc layers are based on the presence and/or signiﬁcance of the terms of equation (3.13) in a particular ﬂow region and cover the whole range of possible ﬂow submergence H m =D. The above ﬂow subdivision and ﬂow types represent a useful schematisation that may help in various problems of gravel-bed ﬂows. For each ﬂow type, a speciﬁc set of relationships describing ‘double-averaged’ ﬂow properties may be developed. The next section addresses the vertical distribution of the double-averaged longitudinal velocity h¯ui, partly based on Nikora et al. (2004).

4.2.

Velocity distribution

Vertical distributions of the time-averaged velocity in gravel-bed ﬂows can be highly variable within a reach and therefore their parameterisation and prediction may be achievable only for ﬂows with high relative submergence and away from the bed where local effects of roughness elements are not felt. Another difﬁculty arises from the fact that most applications deal with spatially averaged roughness parameters that cannot be linked explicitly with local time-averaged velocities, which are variable in space. An alternative approach is to use the double-averaged velocities instead. Their vertical distribution should depend on the ﬂow type and, furthermore, within a particular ﬂow type it may differ in shape from layer to layer (Fig. 3.3). This section ﬁrst brieﬂy considers the least studied ﬂow region deﬁned in Fig. 3.3 as the interfacial sublayer, and then provides some discussion on the velocity distribution above the gravel tops. Considering the simplest case of two-dimensional, steady, uniform, spatially averaged ﬂow over a ﬁxed rough bed, Nikora et al. (2004) derived several models applicable to a range of ﬂow conditions and roughness types that share some common features. Two of these models for the interfacial sublayer (linear and exponential) are directly applicable to gravel bed ﬂows. The exponential model applies

Hydrodynamics of gravel-bed rivers

75

when the effect of the momentum ﬂux downwards dominates over the gravity term in (3.13) that leads to the exponential velocity distribution: h¯uiðzÞ ¼ h¯uiðzc Þ exp bðz zc Þ

(3.14)

where h¯uiðzc Þ is the double-averaged velocity at the roughness crests zc; and b is a parameter. The linear model may be a good approximation for gravel beds where the function f monotonically decreases while the total drag term ðf p þ f n Þ monotonically increases towards the lower boundary of the interfacial sublayer (Nikora et al., 2001), i.e.: h¯uiðzÞ h¯uiðzc Þ ðz zc Þ ¼ u lc

(3.15)

where l c ¼ h¯uiðzc Þ=ðdh¯ui=dzÞzc ¼ dðu =h¯uizc Þ is the shear length scale characterizing ﬂow dynamics within the roughness layer; and d is the thickness of the interfacial sublayer. In principle, relationships (3.14) and (3.15) may be applicable for the interfacial sublayer for all four types of gravel-bed ﬂows deﬁned above, from ﬂows with large relative submergence to ﬂows with partial submergence. Fig. 3.4 supports this conjecture by showing examples of vertical distributions of the double-averaged velocity obtained in laboratory experiments for ﬂow types I, II, and III (high to small relative submergence) and covering roughness types with various densities and arrangements (Nikora et al., 2004). In real gravel-bed rivers, the double-averaged velocity proﬁles within the interfacial sublayer are expected to be more complicated and composed of a combination of the linear and exponential models. The distribution of the double-averaged velocity above the roughness layer for ﬂow type I (with large relative submergence) follows the logarithmic formula (e.g., Nikora et al., 2001): h¯ui 1 zd 1 zd þ C ¼ ln for z zR ¼ ln (3.16) u k dR k z0 where k is the von Karman constant; dR is the thickness of the roughness layer; d is the displacement length (also known as a zero-plane displacement) that deﬁnes the Y=X

[(z)-(zc )]/u*

6 4 2 0 -2 -4 -6 -8 -10

Bead-covered beds Cube-covered beds Spherical-segment bed Gravel-covered beds Fixed 2-d bedforms Rock-type bed 2-d triangle bars Crushed stones

Interfacial sublayer Flow region above the interfacial sublayer

-10

-8

-6

-4

-2

0 2 (z-zc)/lc

4

6

8

10

Figure 3.4. Vertical distribution of the double-averaged velocities for various roughness types in coordinates ½h¯uiðzÞ h¯uiðzc Þ=u ¼ f ½ðz zc Þ=l c (data are fully described in Nikora et al., 2004). Deviations of the data points from Y ¼ X are consistent with the exponential distribution (3.14).

76

V. Nikora

‘hydrodynamic’ bed origin; z0 ¼ dR expðkCÞ is the hydrodynamic roughness length; and the constant C depends on the deﬁnition of dR and the roughness geometry (see Fig. 3.3 for deﬁnitions). It is useful to recall that equation (3.16) has been phenomenologically justiﬁed only for ﬂows with large relative submergence where roughness scale D is well separated from the external ﬂow scale such as mean ﬂow depth H. For the genuine universal logarithmic layer (3.16) to form the required ratio H=D should well exceed 40 or even 80 (Jime´nez, 2004). However, many gravel-bed ﬂows can often be deﬁned as ﬂows with intermediate submergence (ﬂow type II), i.e., they are relatively shallow with respect to the multi-scale bed roughness (H/Do80). With no alternative rigorous theory available these low-submergence ﬂows are nearly always studied using the logarithmic boundary layer concept, which is currently justiﬁed only for deep ﬂows, i.e., H/D480 (Jime´nez, 2004). Nevertheless, the data available for ﬂow type II suggest that the shape of the distribution of the doubleaveraged velocities above the roughness layer is often logarithmic (e.g., Bayazit, 1976; Dittrich and Koll, 1997; Dancey et al., 2003) and, therefore, there might be some general law behind it. Below, an explanation for logarithmic behaviour in ﬂows of type II is suggested by modifying an overlap-based derivation of the logarithmic formula. Following the conventional dimensional analysis we can express the vertical distribution of the double-averaged velocity in the near-bed region above the roughness layer as: h¯ui zd H (3.17) ¼F ; ;g u D D i where gi are the dimensionless parameters of bed roughness (e.g., density of roughness elements); and the ﬂow depth is deﬁned as the difference between the mean water surface elevation and the mean bed elevation. Formula (3.17) represents the inner layer where roughness effects on the velocity ﬁeld dominate. In the ﬂow region well away from the bed, the velocity deﬁcit h¯ui h¯uimax can be expressed as: h¯ui h¯uimax zd H ¼G ; (3.18) u H D where h¯uimax is the maximum ﬂow velocity at z ¼ zm that often occurs at the water surface. Formula (3.18) represents the outer layer where effects of large eddies scaled with the ﬂow depth dominate. Equations (3.17) and (3.18) differ from conventional relationships (e.g., Raupach et al., 1991) by additional variable H=D included in both functions F and G. At very large relative submergence H=D it is reasonable to assume that functions F and G do not depend on H=D. This case corresponds to conventional formulation (3.16) for ﬂows with large relative submergence. However, at smaller values of H=D (ﬂow type II) equations (3.17) and (3.18) state that effects of large eddies on the inner layer are not negligible with, at the same time, bed roughness effects extending into the outer layer. For this case, we can assume that there is an overlap region between the inner and outer layers where equations (3.17) and (3.18) are simultaneously valid, similar to the classical overlap approach. Then, equating the derivatives of (3.17) and (3.18) and multiplying them by (zd) we

Hydrodynamics of gravel-bed rivers

77

can obtain: 1 z d @h¯ui @F @G H H ¼ ZH ¼f ¼ ZD ¼ k u @z @Z D @Z H D D

(3.19)

where Z D ¼ ðz dÞ=D, Z H ¼ ðz dÞ=H, and the function f depends on the relative submergence H=D as it is present as a variable in both F and G. This function is expressed here as f ðH=DÞ ¼ ½kðH=DÞ1 for convenience, preserving classical formulation but interpreting k as the von Karman parameter rather than a constant. From equation (3.19) it is clear that when H=D is not large enough then the von Karman parameter k depends on H=D. Integration of equation (3.19) gives: h¯ui 1 zd H (3.20) ln þC ;g ¼ u kðH=DÞ D D i and h¯ui h¯uimax 1 zd ¼ ln u kðH=DÞ zm d

(3.21)

Equations (3.20) and (3.21) can serve, at least as a ﬁrst approximation, for describing velocity distribution above roughness tops in ﬂows with intermediate submergence (ﬂow type II). The data available for such ﬂows (e.g., Bayazit, 1976; Dittrich and Koll, 1997; Dancey et al., 2003) support equations (3.20) and (3.21) and show that with increase in relative submergence the von Karman parameter kðH=DÞ tends to the well-known universal constant of 0.41. Although relationships (3.20) and (3.21) are to be yet properly tested they represent a useful framework for interpreting and explaining experimental data on ﬂows of type II. 4.3.

Bed origin and zero-plane displacement

Equations (3.16)–(3.21) contain a displacement height d that determines the origin of the logarithmic velocity proﬁle. It is useful therefore to discuss what bed elevation should be used as the bed origin for hydrodynamic considerations. In general, at least three different ‘hydrodynamic’ bed origins may be distinguished, depending on the particular task (Nikora et al., 2002). The bed origin of type 1 corresponds to the level that should be used to measure the ﬂow depth and the bed shear stress. It can be readily shown, using the spatial averaging approach (Nikora et al., 2001), that the spatially averaged bed elevation should be used as the bed origin in this case. The size of the spatial averaging area, which is in the plane parallel to the mean bed and which is the same for both bed topography and the hydrodynamic variables, depends on the statistical structure of bed elevations, i.e., on their probability distribution and the spectrum or correlation functions. The bed origin of type 1 is a natural choice when one considers 2D vertically averaged hydrodynamic equations (models). The deﬁnition of this bed origin does not require any details of velocity distribution as it is based purely on bulk mass conservation and spatially averaged momentum balance. Although the bed origin of this type is equally useful in 3D considerations too, there are also other options for this case. One of them is the bed origin for the logarithmic

V. Nikora

78

formula of the velocity distribution, deﬁned here as the bed origin of type 2. Another useful bed origin, type 3, may be deﬁned for the spatially averaged velocity distribution within the roughness layer. Two natural choices for this bed origin may be considered, i.e.: (i) the minimum elevation of the roughness troughs, and (ii) the upper boundary of the interfacial sublayer as expressed in equations (3.14) and (3.15). Both lower and upper boundaries of the interfacial sublayer may be deﬁned in a statistical sense (e.g., 5 and 95% probability of exceedence). Bed origins of types 1, 2, and 3 do not necessarily coincide as sometimes assumed in research papers. They also do not exclude each other as they represent different ﬂow features and, thus, all of them are useful in modelling and physical considerations. Justiﬁcation for the bed origin types 1 and 3 is clear and reasonably straightforward. However, the bed origin of type 2 for the logarithmic formula is still under debate (see, e.g., Nikora et al., 2002 for review). In Nikora et al. (2002) it was suggested that the zero-plane displacement d for the logarithmic formula should be the level that large-scale turbulent eddies feel as the ‘bed’ and, thus, their dimensions linearly scale with the distance from this virtual bed. Such a deﬁnition directly follows from a slightly modiﬁed Prandtl’s mixing length phenomenology, and serves as a physical basis for determining d from velocity measurements. Nikora et al. (2002) demonstrated that for a range of roughness types the displacement height for the logarithmic formula is strongly correlated with the thickness of the interfacial sublayer and with the shear length scale l c ¼ h¯uiðzc Þ=ðdh¯ui=dzÞzc in (3.15). 4.4.

Fluid stresses in gravel-bed flows

Within the double-averaging framework, the total ﬂuid stress in gravel-bed ﬂows consists of three components:

1 dfh¯ui ~ n hu0 w0 i hu~ wi (3.22) tðzÞ ¼ r f dz which are expressed in equation (3.22), for simplicity, for the case of 2D ﬂows. In most cases the viscous component in equation (3.22) can be neglected, as turbulent stresses in gravel-bed ﬂows are normally several orders of magnitude larger. The spatially averaged turbulent stress hu0 w0 i often changes (quasi) linearly towards the bed and attains a maximum near the roughness tops. Below the roughness tops it reduces to zero due to momentum sink through viscous (skin) friction and form drag. Its distribution is reasonably well studied experimentally showing dependence on relative submergence and roughness geometry. The last component, form-induced ~ appears as a result of spatial correlation of perturbations in timestress hu~ wi, averaged velocities leading to an additional ‘canal’ of momentum ﬂux. The change in ﬂow submergence may lead, in principle, to the change in nature of ﬂuid stresses. Gimenez-Curto and Corniero Lera (1996) suggested that with a decrease in ﬂow submergence, the form-induced stress might become the dominant component of the total stress. This, in turn, may lead to a new ﬂow regime named by the authors the ‘‘jet regime’’, in addition to the well-known laminar, turbulent hydraulically smooth, and turbulent hydraulically rough regimes (Gimenez-Curto and Corniero Lera,

Hydrodynamics of gravel-bed rivers

79

1996). However, the nature of form-induced stresses in gravel-bed ﬂows is still un~ may not be negclear. Some preliminary experimental results suggest that hu~ wi ligible within the roughness layer and their contributions and role in the momentum balance can be important (up to 15–30% of the total ﬂuid stress, Nikora et al., 2001; Aberle and Koll, 2004; Campbell et al., 2005).

5.

Conclusions

In this paper, several issues of gravel-bed river hydrodynamics were discussed, with the focus on two key interlinked topics: velocity spectra and hydrodynamic equations, related to each other through the scale of consideration. It is suggested that the currently used three-range spectral model for gravel-bed rivers should be further reﬁned by adding an additional range, leading to a model that consists of four ranges of scales with different spectral behaviour. This model is considered as a ﬁrst approximation that needs further experimental support. Another topic relates to the spatial averaging concept in hydraulics of gravel-bed ﬂows that provides doubleaveraged transport equations for ﬂuid momentum (and higher statistical moments), passive substances, and suspended sediments. The double-averaged hydrodynamic equations considered in this paper may help in developing numerical models, designing and interpreting laboratory, ﬁeld, and numerical experiments.

Acknowledgements The research was partly funded by the Foundation for Research Science and Technology (C01X0307 and C01X0308), the Marsden Fund (UOA220, LCR203) administered by the New Zealand Royal Society (New Zealand), and the University of Aberdeen (Scotland). Some of studies reviewed in this paper have been completed and published in cooperation with J. Aberle, S. Coleman, A. Dittrich, D. Goring, K. Koll, I. McEwan, S. McLean, and D. Pokrajac, to whom the author is grateful. The author is also grateful for useful discussions and suggestions to J. Aberle, S. Coleman, W. Czernuszenko, J. Finnigan, D. Goring, G. Katul, K. Koll, V.C. Patel, D. Pokrajac, M. Raupach, P. Rowinski, R. Spigel, and R. Walters. R. Spigel and two anonymous reviewers provided useful comments and suggestions which were gratefully incorporated into the ﬁnal manuscript.

References Aberle, J., Koll, K., 2004. Double-averaged ﬂow ﬁeld over static armor layer. In: Greco, M., Carravetta, A., and Della Morte, R. (Eds), River ﬂow 2004, Proceedings of 2nd International Conference on Fluvial Hydraulics, June, 2004, Napoli, Italy, pp. 225–233. Barenblatt, G.I., 1995. Scaling Phenomena in Fluid Mechanics. Cambridge University Press, Cambridge. Barenblatt, G.I., 2003. Scaling. Cambridge University Press, Cambridge.

80

V. Nikora

Bayazit, M., 1976. Free surface ﬂow in a channel of large relative roughness. J. Hydraul. Res. IAHR 14 (2), 115–126. Campbell, L., McEwan, I., Nikora, V., Pokrajac, D., Gallagher, M., 2005. Bed-load effects on hydrodynamics of rough-bed open-channel ﬂows. J. Hydraul. Eng. ASCE 131 (7), 576–585. Dancey, C.L., Kanellopoulos, P., Diplas, P., 2003. Velocity proﬁles in shallow ﬂows over fully rough boundaries. In: Nezu, I. and Kotsovinos, N. (Eds), Inland Waters: Research, Engineering and Management. XXX IAHR Congress, Greece, pp. 119–126. Dittrich, A., Koll, K., 1997. Velocity ﬁeld and resistance of ﬂow over rough surface with large and small relative submergence. Int. J. Sed. Res. 12 (3), 21–33. Finnigan, J.J., 1985. Turbulent transport in ﬂexible plant canopies. In: Hutchinson, B.A. and Hicks, B.B. (Eds), The Forest–Atmosphere Interactions. D. Reidel Publishing Company, Dordrecht, The Netherlands, pp. 443–480. Finnigan, J.J., 2000. Turbulence in plant canopies. Annu. Rev. Fluid Mech. 32, 519–571. Gimenez-Curto, L.A., Corniero Lera, M.A., 1996. Oscillating turbulent ﬂow over very rough surfaces. J. Geophys. Res. 101 (C9), 20745–20758. Gray, W.G., 1975. A derivation of the equation for multi-phase transport. Chem. Eng. Sci. 30, 229–233. Grinvald, D.I., Nikora, V.I., 1988. River Turbulence (in Russian). Hydrometeoizdat, Leningrad, Russia. Jime´nez, J., 2004. Turbulent ﬂows over rough walls. Ann. Rev. Fluid Mech. 36, 173–196. Katul, G., Chu, C.-R., 1998. A theoretical and experimental investigation of energy containing scales in the dynamic sublayer of boundary-layer ﬂows. Boundary-Layer Meteorol. 86, 279–312. Kirkbride, A.D., Ferguson, R.I., 1995. Turbulent ﬂow structure in a gravel-bed river: Markov chain analysis of the ﬂuctuating velocity proﬁle. Earth Surf. Process. Landf. 20, 721–733. Lamarre, H., Roy, A., 2005. Reach scale variability of turbulent ﬂow characteristics in a gravel-bed river. Geomorphology 68 (1–2), 95–113. Livesey, J.R., Bennett, S., Ashworth, P.J., Best, J.L., 1998. Flow structure, sediment transport, and bedform dynamics for a bimodal sediment mixture. In: Klingeman, P.C., Beschta, R.L., Komar, P.D., and Bradley, J.B. (Eds), Gravel-bed Rivers in the Environment. Water Resources Publications, Colorado, pp. 149–176. Lopez, F., Garcia, M., 1996. Turbulence structure in cobbled-bed open-channel ﬂow. Hydraulic Engineering Series No 52, Department of Civil Engineering, University of Illinois at Urbana-Champaign, Illinois. Lopez, F., Garcia, M., 2001. Mean ﬂow and turbulence structure of open-channel ﬂow through emergent vegetation. J. Hydraul. Eng. ASCE 127 (5), 392–402. McLean, S.R., Wolfe, S.R., Nelson, J.M., 1999. Spatially averaged ﬂow over a wavy boundary revisited. J. Geophys. Res. 104 (C7), 15743–15753. Monin, A.S., Yaglom, A.M., 1971. Statistical Fluid Mechanics: Mechanics of Turbulence. Vol. 1. MIT Press, Boston, MA. Monin, A.S., Yaglom, A.M., 1975. Statistical Fluid Mechanics: Mechanics of Turbulence. Vol. 2. MIT Press, Boston, MA. Nelson, J.M., Schmeeckle, M.W., Shreve, R.L., 2001. Turbulence and particle entrainment. In: Mosley, M.P. (Ed.), Gravel-Bed Rivers V. New Zealand Hydrological Society, Wellington, New Zealand, pp. 221–248. Nezu, I., Nakagawa, H., 1993. Turbulence in Open-Channel Flows. A.A. Balkema, Rotterdam, Brookﬁeld, Netherlands. Nikora, V., 2005. Flow turbulence over mobile gravel bed: spectral scaling and coherent structures. Acta Geoph. 53 (4), 539–552. Nikora, V., Koll, K., McEwan, I., McLean, S., Dittrich, A., 2004. Velocity distribution in the roughness layer of rough-bed ﬂows. J. Hydraul. Eng. ASCE 130 (7), 1036–1042. Nikora, V., Koll, K., McLean, S., Dittrich, A., Aberle, J., 2002. Zero-plane displacement for rough-bed open-channel ﬂows. In: Bousmar, D. and Zech, Y. (Eds), Proceedings of the International Conference on Fluvial Hydraulics River Flow 2002, September 4–6, 2002, Louvain-la-Neuve, Belgium, pp. 83–92. Nikora, V., McEwan, I., McLean, S., Coleman, S., Pokrajac, D., Walters, R., 2007a. Double averaging concept for rough-bed open-channel and overland ﬂows: theoretical background. J. Hydraul. Eng. ASCE 133 (8), 873–883.

Hydrodynamics of gravel-bed rivers

81

Nikora, V., McLean, S., Coleman, S., Pokrajac, D., McEwan, I., Campbell, L., Aberle, J., Clunie, D., Koll, K., 2007b. Double averaging concept for rough-bed open-channel and overland ﬂows: applications. J. Hydraul. Eng. ASCE 133 (8), 884–895. Nikora, V.I., 1999. Origin of the ‘‘-1’’ spectral law in wall-bounded turbulence. Phys. Rev. Lett. 83, 734–737. Nikora, V.I., Goring, D.G., 2000a. Eddy convection velocity and Taylor’s hypothesis of ‘frozen’ turbulence in a rough-bed open-channel ﬂow. J. Hydrosci. Hydraul. Eng. JSCE 18 (2), 75–91. Nikora, V.I., Goring, D.G., 2000b. Flow turbulence over ﬁxed and weakly mobile gravel beds. J. Hydraul. Eng. ASCE 126 (9), 679–690. Nikora, V.I., Goring, D.G., McEwan, I., Grifﬁths, G., 2001. Spatially-averaged open-channel ﬂow over a rough bed. J. Hydraul. Eng. ASCE 127 (2), 123–133. Nikora, V.I., Smart, G.M., 1997. Turbulence characteristics of New Zealand gravel-bed rivers. J. Hydraul. Eng. ASCE 123 (9), 764–773. Raupach, M.R., Antonia, R.A., Rajagopalan, S., 1991. Rough-wall turbulent boundary layers. Appl. Mech. Rev. 44 (1), 1–25. Raupach, M.R., Shaw, R.H., 1982. Averaging procedures for ﬂow within vegetation canopies. BoundaryLayer Meteorol. 22, 79–90. Roy, A.G., Bufﬁn-Belanger, T., 2001. Advances in the study of turbulent ﬂow structures in gravel-bed rivers. In: Mosley, M.P. (Ed.), Gravel-Bed Rivers V. New Zealand Hydrological Society, Wellington, New Zealand, pp. 375–404. Roy, A.G., Bufﬁn-Belanger, T., Lamarre, H., Kirkbride, A.D., 2004. Size, shape, and dynamics of largescale turbulent ﬂow structures in a gravel-bed river. J. Fluid Mech. 500, 1–27. Townsend, A.A., 1976. The Structure of Turbulent Shear Flow. Cambridge University Press, Cambridge. Whitaker, S., 1999. The Method of Volume Averaging. Kluwer Academic Publishers, Dordrecht. Wilcock, P.R., 2001. The ﬂow, the bed, and the transport: interaction in ﬂume and ﬁeld. In: Mosley, M.P. (Ed.), Gravel-Bed Rivers V. New Zealand Hydrological Society, Wellington, New Zealand, pp. 183–220. Wilson, N.R., Shaw, R.H., 1977. A higher order closure model for canopy ﬂow. J. Appl. Meteorol. 16, 1197–1205. Yaglom, A., 1993. Similarity laws for wall turbulent ﬂows: their limitations and generalizations. In: Dracos, Th. and Tsinober, A. (Eds), New Approaches and Concepts in Turbulence. Birkhauser Verlag, Basel, pp. 7–27.

Gravel-Bed Rivers VI: From Process Understanding to River Restoration H. Habersack, H. Pie´gay, M. Rinaldi, Editors r 2008 Elsevier B.V. All rights reserved.

85

4 Pressure- and velocity-measurements above and within a porous gravel bed at the threshold of stability Martin Detert, Michael Klar, Thomas Wenka and Gerhard H. Jirka

Abstract Experimental results of measurements characterising the pressure and velocity above and within a porous gravel layer are presented. The goal of this study is to give a better understanding of the ﬂow in the hyporheic interstitial under the inﬂuence of turbulence in the main ﬂow. Latest developments in measuring techniques were applied: miniaturised piezoelectric pressure sensors (MPPS) were used to measure turbulent pressure ﬂuctuations inside the gravel layer. Velocity measurements were carried out by a 3D-particle tracking velocimetry system (3D-PTV) using miniaturised endoscopic stereo setups within artiﬁcial gravel pores. Additionally, in the main ﬂow a 1D-acoustic doppler current proﬁler (1D-ADCP) was used. Within the main ﬂow, alternating faster and slower ﬂuid packets with a size scaling with the ﬂow depth were observed. Pressure ﬂuctuations rms( p) as well as velocity ﬂuctuations rms(u, v, w) decrease exponentially with increasing gravel depth, mainly within the ﬁrst two layers of gravel grains. 1.

Introduction

Many problems in hydraulic engineering in rivers and waterways are related to the prediction of the morphodynamical development of the bed. The efﬁciency of regulation works or hydraulic constructions such as groynes, weirs or embankments is strongly inﬂuenced by the stability of the river bed and the artiﬁcial geotechnical armoring layer, respectively. For example, 170,000 m3 of gravel feeding per year is needed for the Iffezheim barrage at the Rhine River near Karlsruhe to avoid erosion in the downstream river bed. The costs run up to 5 Mio. Euro each year (WSA, 2004). Over the last 100 years much research work has been done to gain insight into the background of river bed stability. Shields (1936) developed a concept of a critical shear stress t0c to describe the transition from a stable to a moving bed. This can be seen as the ‘classical’ approach in bed stability. Diverse formulae have been E-mail address: [email protected] (M. Detert) ISSN: 0928-2025

DOI: 10.1016/S0928-2025(07)11121-4

M. Detert et al.

86

developed to improve this approach, e.g. Zanke (2001). Furthermore, a large number of mainly empirical approaches were published to predict erosion in nonuniform ﬂow regions like piers, groyne heads or due to jets, etc. However, up to now no satisfactory, physically founded description has been established to answer deﬁnitely the question of bed stability. The objective of this paper is to improve the physical understanding of the hydrodynamic process above and within gravel beds. The long-term goal is to better understand erosion and sedimentation as well as exchange processes (mass, momentum) between surface and subsurface water. To study the ﬂow within the hyporheic interstitial, miniaturised piezoelectric pressure sensors (MPPS) and 3D-particle tracking velocimetry system (3D-PTV) were applied.

2.

Background

Instantaneous shear and pressure forces acting on outer grains are the driving mechanisms for destabilisation processes of a mobile bed. Beyond their inﬂuence on the stability of a river bed, turbulent structures are also a determining factor for colmation processes and mass transfer in the transition between main ﬂow and subsurface interstitial ﬂow (e.g., Vollmer et al., 2002; Roy et al., 2004). To describe these mechanisms there is a need for synoptic measurements of velocity and pressure ﬁelds in the main ﬂow as well as within the pore ﬂow within the bed that is interacting with the main ﬂow. Dynamic processes have a relevant inﬂuence on the ﬂow near a rough porous bed. Eddies emerge from wakes behind roughness elements and rollup processes at free shear layers. The latter eddy-generating mechanism was ﬁrst described for smooth beds by Kline et al. (1967) as the bursting phenomenon. It generated a new interest in studying the structures of boundary layer turbulence (Grass, 1971; Dittrich et al., 1996; Sechet and Guennec, 1999; Adrian et al., 2000). These coherent turbulent structures play an important role in the pressure peaks acting on the bed. Farabee and Casarella (1991) derived from literature and own wind tunnel experiments, that the relation between the pressure variance rms(p) and the bed shear stress t0 depends on the Reynolds number for the boundary layer thickness d and shear velocity u*, Re*, d ¼ u*d/V. The measured range is qﬃﬃﬃﬃﬃ p2 rmsðpÞ ¼ ¼ 2:5 3:5 (4.1) t0 t0 With the estimate of max(p)/rms(p) ¼ 6, found by Emmerling (1973), the maximum pressure peaks can reach up to max(p) ¼ 18 t0. Thus, loads from max(p) appear to be more than one order of magnitude larger than the critical Shields parameter t0c. In contrast to the knowledge concerning the ﬂow over a smooth bed, the research concerning the ﬂow over and within a rough porous bed cannot be considered complete. Twenty years after his ﬁrst works, Grass et al. (1991) showed that coherent structures are also detectable on rough beds. Likewise, Smith et al. (1991) and Deﬁna (1996) found that the lateral extension of these eddy structures merely correlates with

Pressure- and velocity-measurements

87

the roughness height. Garcı´ a et al. (1996) depicted the effect of the coherent structures on the entrainment and transport of particles in a ﬂow over smooth and over rough beds. Sechet and Guennec (1999) showed that the high energy containing turbulent structures govern the Reynolds stresses signiﬁcantly. By means of experimental data they were able to correlate sediment transport with burst-like eddy structures near the bed. Hoﬂand (2004) and Hoﬂand et al. (2005) used a 2D-PIV technique and three piezometric pressure sensors on top of a rough bed of stones at the threshold to instability. They were able to correlate the movement of a single stone to a mechanism of small-scale lift ﬂuctuations followed by a large-scale drag force. The depiction of ﬂow on top of a rough bed from Hoﬂand et al. (2005) is the most qualitative at present. However, as they focus on the uppermost gravel layer, no information about the interaction between open-channel and pore ﬂow can be given. To improve the basic knowledge of this interaction, both the surface and the interstitial ﬂow have to be surveyed.

3. 3.1.

Experiments Experimental setup

The experimental system was implemented in an open-channel ﬂume located at the Federal Waterways Engineering and Research Institute (BAW), in Karlsruhe. The ﬂume was L ¼ 40 m long and B ¼ 0.9 m wide. The maximum ﬂow rate Qmax ¼ 0.275 m3/s led to low mobility conditions at the bed. In the ﬂume (see Fig. 4.1), a sand layer of HS ¼ 0.5 m was covered by a porous gravel layer of HP ¼ 0.10 m. The effective length of the ﬂume was L ¼ 30.10 m. The measurement area was located in the middle

(a)

Streamwise view

(b) Side view

Figure 4.1. Sketch of experimental setup with deﬁnition of the coordinate system, dimensions in [m], not to scale. (a) streamwise view with the positions of the MPPSs, (b) side view, where the positions of the three artiﬁcial gravel pores with 3D-PTV and the MPPSs are shown.

M. Detert et al.

88

of the ﬂume, hence inﬂuences of inlet (fully developed boundary layer) and outlet were negligible. The medium grain diameter of the gravel was dmD ¼ 10.2 mm, with a degree of nonuniformity of Cc ¼ d60/d10 ¼ 1.25. Thus, the gravel was very uniform. The pore number was determined to be e ¼ 62.4% (loose bulk density: rbulk ¼ 1.538 g/cm3, absolute density: rabs ¼ 2.464 g/cm3). The mean grain diameter of the uniform sand was about ds ¼ 1.0 mm. The critical shear stress t0c for this material is calculated to t0c ¼ 8.8 Pa, using Shields equation with Frc* ¼ t0c/(Dr g dmD) ¼ 0.06. Fig. 4.1 illustrates the measurement setup. In particular, the setup consists of the following parts:

three endoscopic stereo probes to record image sequences of the interstitial ﬂow inside the gravel layer with specially prepared artiﬁcial gravel pores; ﬂow analysis by a 3D-PTV algorithm. up to ten MPPS at arbitrary locations within the gravel layer; three of the sensors attached to the artiﬁcial gravel pores.

To gain additional insight into the velocity regime, a 1D-acoustic doppler current proﬁler (1D-ADCP) was applied. In the following sections the measuring techniques are described in more detail.

3.2.

3D-particle tracking velocimetry (3D-PTV)

The 3D-PTV technique is a nonintrusive optical technique based on image sequence analysis. Pore ﬂow measurements inside three single pores of the gravel layer were carried out using three miniaturised endoscopic stereo setups. The basic principle of these setups was to acquire stereoscopic image sequences of the ﬂow ﬁeld inside the pore volume by viewing it from two different directions. Two ﬂexible ﬁberoptic endoscopes of 2.4 mm diameter were attached to an adapted artiﬁcial gravel pore made of grains ﬁxed to each other (Fig. 4.2). The optical axes of the two cameras enclosed an illumination fiber pressure sensor

pore volume

holes for fixation of stereo rig

(a)

(b)

Figure 4.2. Endoscope stereo setup. (a) For velocity measurements in the gravel layer, this stereo rig is attached to an artiﬁcial gravel pore, viewing the pore volume inside. (b) Artiﬁcial gravel pore.

Pressure- and velocity-measurements

89

angle of about 901. The size of the stereo volume was about 5 mm in all directions. For more details see Klar et al. (2002). Illumination of the pore volume was provided by an optical ﬁberbundle guiding the light from a halogen cold light source into the pore. The purpose of the artiﬁcial pore is to hold the endoscopes and the illumination ﬁber at a ﬁxed relative position and to keep surrounding grains in the gravel layer from blocking the endoscope view. To perform ﬂow measurements, the three artiﬁcial pores were embedded in the gravel layer at different positions (see again Fig. 4.1). A suspension of tracer particles was added to the ﬂow upstream of the pores, and particle image sequences of the two different endoscope views were recorded simultaneously. Since the tracer particles cover this small volume very rapidly, the cameras must operate at high frame rates. Hence, two of the three artiﬁcial pores were equipped with highspeed MegapixelCMOS-cameras (Photonfocus MV-D1024). The read-out size of the cameras had been set to 184 184 pixels. Zoom optics were used to ﬁt the endoscope image to this image size. By decreasing the image resolution in this way, a maximum frame rate of 400 Hz can be achieved. The third endoscope setup was working with standard CCD cameras running at 50 Hz. This setup could only be used in the lowermost position within the gravel layer, where the ﬂow velocities and ﬂuctuations are expected to be lowest. The image data of all three setups were written to RAID hard-disk arrays in real-time during the acquisition. Thus, the duration of the sequences was only limited by the RAID capacity. For a single measurement, a sequence duration of Dt ¼ 60 s has been chosen. In order to extract 3D velocity information from the image data, the stereo sequences are processed by a 3D-PTV algorithm. The result is a set of 3D Lagrangian ﬂow trajectories of the tracer particles suspended in the water. In the 3D-PTV method, the underlying real velocity ﬁeld is sampled at random points both in space and time. Velocity information is available only at those positions and time instants where tracer particles could be found and successfully tracked. If the ﬂow is seeded homogeneously, this is not a severe limitation. Note that in PIV the distribution of tracers is also random, but the density is high and homogeneous enough to enable the determination of velocity vectors on a regular grid. In the experiments presented here, it was not possible to seed the whole water volume with tracer particles due to the very large size of the ﬂume and the water supply system, which was also connected to several other experiments. Thus, tracers had to be added to the pore ﬂow punctually in the vicinity of the artiﬁcial pores. With this seeding method, a homogeneous tracer distribution could not always be obtained.

3.3.

Pressure measurements by MPPS

The MPPSs were developed to measure pressure ﬂuctuations within and on top of the gravel layer. This insitu technique turned out to be a very robust and reliable tool to determine pressure ﬂuctuations down to dissipative scales. They offer a high adaptability within a hostile, rough environment by water resistent housing and ﬂexible cables. The principle of the MPPS is based on the piezoresistive effect. The initial point is an element of silicium, with implanted resistances in its bending panel. Fig. 4.3

M. Detert et al.

90

(a)

(b)

Figure 4.3. (a) Head of MPPS outside the ﬂume. (b) Head of MPPS ﬁxed at y ¼ 10 mm above the gravel layer, facing upstream.

shows two photographs of the encapsulated head of the pressure pickup. The sensors were locally ﬁxed on a grid to keep them on an accurately deﬁned position. The differential pressure is measured in reference to atmospheric pressure, with compensation of temperature. The sensors are encapsulated with slowly hardening epoxy resin and sealed up with clear varnish to make them water resistent. The maximal dimensions of the sensors are 2 1.2 1.2 cm3, with a shape similar to a gravel grain of the larger size fraction. Due to signal conditioning by the purpose-built ampliﬁer the guaranteed maximum measurable frequency is 100 Hz. To avoid aliasing effects, the measurements are recorded at a rate of 500 Hz. With a tolerance in accuracy of less than 1.0% full scale, the encapsulated sensors were point-calibrated at 5 V according to 3 kPa and 10 V according to 6 kPa, respectively. Hence, the absolute range of the pressure sensors is 0–6 kPa which equals 0–587.4 mmWC at 201C. The accuracy of the 12-bit AD card is limited to 587.4/212 ¼ 0.14 mmWC. However, it was possible to improve the resolution by utilisation of the dithering effect and ﬁltering techniques to40.003 mmWC for fo20 Hz and40.012 mmWC for f420 Hz, respectively (Detert et al., 2004). Measurements were performed simultaneously by up to 10 pressure sensors over 2 min. Pressure sensors were located at vertical positions of y/dmD ¼ 1.0 (above), 0.0 (at top) and at various positions within the gravel layer. On each of the three artiﬁcial gravel pores a sensor was adapted to gain simultaneous insight in pressure and velocity, respectively. 3.4.

Velocity measurements with 1D-ADCP

An additional insight into the velocity regime of the open-channel ﬂow was gained by an 1D-ADCP, namely a DOP 1000 (Willemetz, 1997). In pulsed Doppler ultrasound, instead of emitting continuous ultrasonic waves, an emitter periodically sends out a short ultrasonic burst and a receiver continuously collects the echo issues from targets that may be present in the path of the ultrasonic beam. By sampling the incoming echoes at the same time relative to the emission of the bursts, the shift of positions of

Pressure- and velocity-measurements

91

scatterers are measured. Velocities are derived from the shifts in positions between the pulses. The operating principle is depicted in Fig. 4.4 for a vertical wall distance of y ¼ 150 mm and a doppler angle of y ¼ 601 against the streamwise direction, as it was installed for most of the experiments. The parameters of the 1D-ADCP have to be adjusted for each ﬂow condition. Thus, within the measurements the resolution in space varied from 1 to 2 mm in beam direction and the resolution in time from 15 to 60 ms. The measuring technique of the 1D-ADCP allows measurement of reliable velocity proﬁles from yZ5 mm above the gravel layer. Closer to the gravel reﬂections of the ultrasonic beam lead to errors in measurement. The maximum detectable velocity in streamwise direction

150

Figure 4.4. Operating principle and geometry of the 1D-ADCP as it was installed during the experiments. The diameter of the acoustic ﬁeld (main lobe) spreads from 5 mm at the transducer to the top edge of the gravel layer with a diameter of 16.7 mm containing 80%, and to 33.5 mm containing 98% of the acoustic ﬁeld intensity, respectively. Dimensions in [mm], not to scale.

M. Detert et al.

92

resulted in max(u) ¼ 1.3 m/s. This was appropriate to the expected velocities. Proﬁles were measured with a duration of 1 min at various distances to the side wall in order to get time-averaged isoline velocity plots. Working with the 1D-ADCP in the measuring area was not possible because of spatial limitations. Therefore this instrument was located at x ¼ 3.3 m downstream the measuring area. By this way the other measuring techniques were not disturbed. The velocity component ux measured by the 1D-ADCP is always the component in the direction of the ultrasonic beam, x. A spacial vectorial decomposition was used to estimate the horizontal velocity component u. Starting from ux ¼ cos y ~ u þ sin y~ v

(4.2a)

! ! ! ux þ ux 0 ¼ cos yð¯u þ u0 Þ þ sin yð¯v þ v0 Þ

(4.2b)

and neglecting secondary ﬂow effects (¯v ﬃ 0), the time average of equation (4.2) simpliﬁes to u¯ ¼

1 ¼ 2:0ux cos y ux

(4.3)

However, an instantaneous analysis of ﬂow structures can only be made along the beam axis, x. 3.5.

Measurement programme

The measurement programme was designed for variation of water depth, thickness of the gravel layer as well as ﬂow conditions up to low mobility conditions. Table 4.1 gives the ﬂow conditions and mean parameters of the experiments A01–A10. Within this series, the bed shear stress t0 was gradually increased to low mobility conditions. A criteria of instability is deﬁned by t0/t0c, with t0c ¼ 8.8 Pa after Shields (1936). At t0/t0c ¼ 0.59 the transport of single grains was observed. This agrees with analytical studies of Dittrich et al. (1996) and with experimental data from Wilcock et al. (1996) for low mobility conditions and loose bed density. Table 4.2 gives the positions of the gravel sensors. In the vertical direction, y ¼ 0 is deﬁned at 0.25 dmD below the uppermost gravel grains. Table 4.1.

Experimental conditions with a gravel layer thickness of HP ¼ 0.10 m.

Series

Unit

A01

A02

A04

A06

A08

A10

t0/t0c Q h U u* Re* ¼ u* dmD/v UADCP hADCP

[] [m3/s] [m] [m/s] [m/s] [] [m/s] [m]

0.09 56.0 0.201 0.31 0.026 260 0.31 0.200

0.18 81.8 0.203 0.45 0.040 410 0.45 0.200

0.36 120.5 0.207 0.65 0.063 640 0.67 0.199

0.48 149.8 0.219 0.76 0.078 800 0.80 0.207

0.55 173.0 0.234 0.82 0.073 740 0.86 0.224

0.59 193.4 0.249 0.86 0.085 870 0.91 0.235

Pressure- and velocity-measurements Table 4.2.

93

Positions of the pressure sensors for runs A01–A10.

Sensor

y (cm) (vertical)

x (cm) (longitudinal)

z (cm) (transversal)

Comment

P+10 P00a P00b P10 P20 P55 P65 P75

1.0 0.0 0.0 1.0 2.0 5.5 6.5 7.5

0.0 1.2 1.2 1.5 26.0 1.5 18.0 1.5

6.0 3.0 9.0 10.5 0.0 0.0 0.0 10.5

Above gravel, see Fig. 4.3(b) Within gravel, facing up Within gravel, facing up Within gravel, facing up At artiﬁcial gravel pore At artiﬁcial gravel pore At artiﬁcial gravel pore Within gravel, facing up

(Eq.5)

(Eq.4b)

(Eq.4a)

Figure 4.5. Velocity proﬁles u¯ =u and derived proﬁles rmsðuÞ=u , measured by 1D-ADCP. In comparison to equations (4.4) and (4.5). Run A01–A10, U ¼ 0.31–0.91 m/s, t0/t0c ¼ 0.090.59, water depth HAﬃ0.20 m.

4. 4.1.

Results Flow above the gravel layer

A depiction of the ﬂow conditions above the gravel layer is given by results from the 1D-ADCP measurements. Nondimensionalised proﬁles of u¯ =u and proﬁles of mean velocity ﬂuctuations rmsðuÞ=u gained by the 1D-ADCP are given in Fig. 4.5. Each proﬁle results from 16 proﬁles evenly spread over the width of the ﬂume, B ¼ 0.90 m, and represents a time average over approximately 1 min. The roughness height was determined by curve ﬁtting to kS ¼ 1.68 dmD with the assumption y ¼ 0 at 0.25 dmD below the uppermost gravel grains. Unfortunately, the determination of the shear velocity, u , is subject to uncertainties. u was determined in two indirect ways. The ﬁrst way was to calculate u* from the measured water surface slope. Especially at U40.70 m/s this method led to problems due to noneven water level. Therefore, u* was

M. Detert et al.

94

determined from roughness parameters that were identiﬁed in calibration tests before. The ultimate remaining error for u* was estimated to 75%. The velocity proﬁles conform to the log law for y/ho0.2 (equation (4a)) as well as to the semiempirical wake-function by Coles (1956) for the outer region y/h40.2 (equation (4b)). uðyÞ 1 y þ 8:48 (4.4a) ¼ ln u 0:4 ks y uðyÞ 1 y þ 8:48 þ 1:25 sin2 p ln ¼ u 0:4 ks 2h

(4.4b)

Problems in ﬁtting the log proﬁles for U ¼ 0.45 m/s (A02) and especially for U ¼ 0.31 m/s (A01) are supposed to result from a lack of particles passing the measuring beam. The lower the ﬂow velocity, the lower is the number of particles that reﬂect the ultrasonic beam. Thus, some ultrasonic bursts are possibly not reﬂected, what might lead to misinterpretation of the recorded signal. The projected proﬁles of the turbulence intensity were determined by the assumption of 2 rmsðux Þﬃrms(u) (see equation (4.3)). They are compared with the semiempirical formula given by Nezu and Nakagawa (1993): rmsðuÞ ¼ 2:30 e1:0y=h u

(4.5)

The order of magnitude of all proﬁles rmsðuÞ=u roughly complies with equation (4.5). But a closer look shows, that the measured rms(u)/u* increase less strongly near the gravel bed. There are two reasons for this deviation. First, eddies that are smaller than the measuring volume are not detectable (biasing effect). Second, near the bed the signal is deteriorated by beam reﬂections at the bed. Both effects result in an underestimation of the real velocity ﬂuctuations. A proper explanation of instantaneous processes can only be given along the ultrasonic beam axis. Fig. 4.6 gives a series of measured instant proﬁles of velocity ﬂuctuations ux 0 ¼ ux ux , here as an example gained from run A06. Alternating faster and slower ﬂuid packets with a size up to half of the water depth, oh/2, are passing the 1D-ADCP beam. The total clip time of this depiction is Dt ¼ 0.791–0.422 ¼ 0.369 s. Note that ux 0 denotes the instantaneous ﬂuctuations in line of the beam axis, x, and not in streamwise direction, x. From t ¼ 0.422–0.580 s, an older slower ﬂuid packet I ðux 0 o0Þ disappears out of the range of the 1D-ADCP and a newer packet II with faster ﬂuid ðux 0 o0Þ becomes signiﬁcant. At time step t ¼ 0.633 s, a slower packet III starts to displace the packet II downwards. At t ¼ 0.685 s, the packet II still exists and reaches its maximal velocity max(ux 0 Þﬃ130 mm=s at y ¼ 40 mm. With ux (y ¼ 40 mm) ¼ 360 mm/s, this corresponds to 131% of ux . Shortly afterwards at t ¼ 0.691 s, packet III is partly displaced downward by a faster ﬂuid packet IV. This observation agrees with the ﬂow pattern model in the outer region of a boundary layer over smooth beds described by Adrian et al. (2000). The model speciﬁes multiple uniform momentum zones that have a characteristic growth angle (mostly 121). These packets consist of eddies that propagate together with the packet convection speed. Unfortunately, the 1D-ADCP is not capable to resolve these ‘footprints’ of smaller

Pressure- and velocity-measurements

95

Figure 4.6. Velocity ﬂuctuations ux 0 ¼ ux u x in direction of the ultrasonic beam, x, measured by 1DADCP. For graphical visualisation ux 0 is projected horizontally. Run A06, UADCP ¼ 0.76 m/s, u x (y ¼ 40 mm) ¼ 360 mm/s, t0/t0c ¼ 0.48, and water depth hADCP ¼ 0.219 m.

M. Detert et al.

96

vortices. Within these limitations, the 1D-ADCP as an acoustic instrument consequently may help to resolve bigger ﬂow patterns and verify multidimensional measuring techniques like 3D-PTV or PIV, respectively. However, the 1D-ADCP does not have the potential to replace the other techniques. 4.2.

Flow within the gravel layer

Fig. 4.7 presents the pressure ﬂuctuations rms(p) measured at various positions above and within the gravel layer as well as with increasing shear stress t0. Both parameters are nondimensionalised by t0c. At t0/t0c ¼ 0.59 low mobility conditions were detected, as single stones passed the measuring area. In order to avoid mechanical deformation of the whole sensible measuring system, larger stress rates were not examined. Within a ﬁrst approximation, all curves in Fig. 4.7 increase linearly with the instability criteria t0/t0c. Focussing on the two sensors P00b and P00a at the interface between gravel and open-channel ﬂow, the inclination can be determined to a ratio of rms(p)/t0 ¼ 1.9/0.6 ¼ 3.2, which agrees with equation (4.1). Furthermore, the damping due to the gravel becomes obvious. At P10 within the gravel the rate is about 1.7/0.6 ¼ 2.8. Deeper in the gravel bed no essential difference between vertical positions can be detected. There the ratio is given mostly by 22.5. The maximum ﬂuctuations within the main ﬂow at y/dmD ¼ 1.0 above the gravel are 3.54 times the medium ﬂuctuations. The pressure sensor P+10 is facing against the streamwise direction (see also Fig. 4.3b). Because of this, the measured signal is inﬂuenced by both p0 and 1/2ru0 2. 8 7 6 horizontal arrangement

rms(p) /τoc [−]

+1.0 cm 5 4

vertical arrangements

3 0 cm 2 1 −2.0 cm 0

0.1

0.2

0.3

τo / τoc [−]

0.4

0.5

−1.0 cm 0.6

Figure 4.7. Pressure ﬂuctuations rms(p) for increasing t0, both normalised by t0c ¼ 8.8 Pa. Runs A01–A10. Positions of the sensors: see Table 4.2.

Pressure- and velocity-measurements

97

150 +1cm

p′(t) [Pa]

100

50

-7.5cm

0

-50 -1cm -100 38

40

42

44

46

48 t [s]

50

52

54

56

58

Figure 4.8. Simultaneous time traces p0 (t) [Pa]. Run A10, U ¼ 0.86 m/s, t0/t0c ¼ 0.59, and water depth h ¼ 0.249 m. Positions of the sensors: see Table 4.2.

In Fig. 4.8, a more detailed view on the pressure ﬂuctuations p0 (t) is presented. The given example is a time clip over Dt ¼ 5838 ¼ 20 s from run A10, i.e. at low mobility conditions. The damping of higher frequencies with increasing gravel depth becomes obvious. A correlation between the pressure signal within low frequencieso1 Hz can be identiﬁed, associated to macro-ﬂuid structures and corresponding water level oscillations. The distances between the pressure sensors were not smaller than 20 mm, due to the size of the sensors. Therefore a correlation for higher frequencies with smaller corresponding length scales cannot be detected. Furthermore, a closer view to Fig. 4.8 shows that results p0 40 for the sensor above the gravel layer occur less frequently, but with more extremal values. Within the main ﬂow the pressure peaks max(p) are predominantly negative and less extreme than the positive peaks. Within the gravel layer this distribution equalises. The histograms in Fig. 4.9 give a more vivid description to this fact. The probability density functions (PDF) given as example, result from measurements shown in Fig. 4.8. The distributions are characterised statistically by the skewness S(p) and the curtosis (or concavity) K(p). SðpÞ ¼

KðpÞ ¼

n 1X p 03 n i¼1 i n 1X p 04 n i¼1 i

(4.6a) ! 3

(4.6b)

The Gaussian normal distribution with S(p) ¼ 0 and K(p) ¼ 0 reads P

p0 rmsðpÞ

¼

2 1 0 pﬃﬃﬃﬃﬃﬃ e1=2ðp =rmsðpÞ0Þ : 1 2p

(4.7)

M. Detert et al.

98 0.45

0.45

vert. arrangement y/dmD = −1.0 rms(p)= 15.9 Pa S(p) = −0.03 K(p) = 0.10

0.4

0.35

0.3

P(p′/rms(p)) [−]

P(p′/rms(p)) [−]

0.35

0.25 0.2

Eq. 7 (Gauss)

0.15

0.3 0.25 0.2

0.1

0.05

0.05

−3

(a)

−2

−1

0

1

p′/rms(p) [−]

2

3

Eq. 8 δ = 3.0 (Hofland)

0.15

0.1

0

horiz. arrangement y/dmD = +1.0 rms(p) = 70.6 Pa S(p) = 0.94 K(p) = 1.09

0.4

0

−3

4

(b)

−2

−1

0

1

2

3

4

p′/rms(p) [−]

Figure 4.9. Histograms P(p0 /rms(p)) at (a) y/dmD ¼ 1.0 in comparison to equation (4.7) and (b) y/dmD ¼ 1.0 in comparison to equation (4.8) (d ¼ 3, 0). Run A10 with Dt ¼ 120 s, U ¼ 0.86 m/s, t0/t0c ¼ 0.59, and water depth h ¼ 0.249 m.

For S(p) ¼ 0.03 and K(p) ¼ 0.10 the PDF at y/dmD ¼ 1.0 within the gravel layer follows this Gaussian shape. The distribution above the gravel at y/dmD ¼ 1.0 has a positive skewness with S(p) ¼ 0.94, and a narrow crested curtosis with K(p) ¼ 1.09. It ﬁts very well with a distribution proposed by Hoﬂand and Battjes (2006), referring to the quadratic approach p|u|u (equation (4.8)). It is related to the w2 distribution, i.e. the PDF of the square of a normally distributed variable. pﬃﬃﬃﬃﬃ 2 ðA þ mÞ s ¼ pﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ e1=2ð jAjsignðAÞdÞ ½ for A_0 P (4.8a) s 2 2pjAj with A¼

s p0 rmsðpÞ

(4.8b)

d¼

Ub rmsðub Þ

(4.8c)

sðdÞ ¼

pﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ 2 4d2 þ 2 þ eð0:55d Þ

mðdÞ ¼ ðd2 þ 1Þ e1:63d

(4.8d) (4.8e)

The noncentrality parameter d (equation (4.8c)) has to be interpreted as the inverse of the relative turbulence intensity rms(ub)/Ub near the bed. The standard deviation s and the mean m can be calculated by ﬁtted functions given by equations 4.8d and e. As shown in Fig. 4.9b, the shape of the PDF is predicted almost perfectly by equation (4.8) for chosen turbulence intensity of d ¼ 3.0. Thus, the pressure ﬂuctuations measured above the gravel layer depend on u0 |u0 |. However, the normal distributed pressure

Pressure- and velocity-measurements

99

ﬂuctuations within the gravel layer are less dependent on the turbulent velocity ﬂuctuations of u0 |u0 |. The dominant mechanism must be another one. The frequency-characteristic of the pressure signals is examined by the Fourier transformation. By means of the Fourier transformation, the pressure signal is transformed from the time domain into the frequency domain, as a decomposition into sine (or cosine, respectively) oscillations. The results are power spectral densities (PSD), where dominating frequency ranges and corresponding energies are revealed. Fig. 4.10 gives an exemplary PSD for low mobility conditions at t0/t0c ¼ 0.59. For the sensors P+10,P00b and P00a above and on top of the gravel layer the results disagree with the classical Kolmogorov k7/3 (f7/3) law for the turbulence cascade in openchannel ﬂow, where k is the wave number and f the frequency (Nezu and Nakagawa, 1993). But the inclination agrees with results presented by Gotoh and Rogallo (1999), where a second inertial k5/3 range at smaller scales is described. Within the gravel

f−7/3

103

f−5/3

PSD [Pa2/Hz]

102

101

Eq. 9 Eq. 10

P+10 100 P−10

P00a P00b

others P−20 10-1 10-1

100

101

102

f [Hz] Figure 4.10. Power spectral density (PSD). Run A10, U ¼ 0.86 m/s, t0/t0c ¼ 0.59, water depth HAﬃ0.20 m.

100

M. Detert et al.

layer a signiﬁcant damping between 1 and 3 Hz can be recognised. Below y/dmD ¼ 2.0 within the gravel layer there is no identiﬁable difference in damping pressure ﬂuctuations higher than 3 Hz. It is hypothesised that the pressure ﬂuctuations are dominated by the long wave ﬂuctuating water level. The inﬂuence of an oscillating water level has been estimated as follows. A wavelength of L ¼ 0.40–2.5 m with a constant small amplitude of a ¼ 0.5 mm is assumed. By negligence of the surface tension, the wave theory of ﬁrst order gives the resulting maximal bed pressure due to surface waves by rga (4.9) maxðp0 ðkÞÞ ¼ coshðkhÞ The wave number is given by k ¼ 2p/L. Within the transit from deep to shallow water, at 0.05oh/Lo0.5, the corresponding wave frequency reads 1 pﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ g=k tanhðkhÞ (4.10) f ¼ L As it can be seen in Fig. 4.10, equation (4.9) combined with equation (4.10) coincides very well with the spectra calculated for the lower part of the gravel layer. Consequently, the long wave ﬂuctuations are strongly inﬂuenced by an oscillating water level. At this moment, it can not be answered deﬁnitely whether these changes in water level are due to the experimental setup (e.g., imperfections in side wall) or due to macroﬂuid structures. There is a need for an intensive comparison to the velocity ﬁelds in the open-channel ﬂow. But referring to Hoﬂand (2002), it can be supposed that these long wave ﬂuctuations play a minor role for the entrainment of single grains. A moving or rolling gravel grain and the driving ﬂuid structure must have the same length scale. However, without multidimensional velocity information above the gravel layer, the correlation of time/frequency scales and length scales is difﬁcult. The simplest way to separate the inﬂuence of an oscillating water level is to subtract the PSD gained from the lowermost sensor P75 from the other PSDs. The remaining signal is purged from water level effects. The PSDs depicted in Fig. 4.11 are simpliﬁed in this way. They are scaled by u eliminate the timescale. The PSDs are roughly congruent, independent if the hydromechanical load is corresponding to stable bed conditions at t0/t0c ¼ 0.18 (A02) and 0.36 (A04) or to low mobility conditions at t0/t0c ¼ 0.59 (A10). The characteristics of frequency and corresponding power are the same. Fig. 4.12 shows the damping rms(p)/t0 with increasing gravel depth. The pressure signal here was ﬁltered with a high-pass f41.5 Hz to eliminate the long waves inﬂuence (see Fig. 4.12). The impact of the outer ﬂow on the pore ﬂow in the upper grain layer becomes clearly apparent, as rms(p) is relatively high. Moreover, all series of experiments show an exponential damping of the pressure ﬂuctuations with an increasing gravel depth. Deeper down a relative layer depth of y/dmDo2.0, the detected ﬂuctuations stay nearly constant, as the ﬂow is mainly dominated by the seepage ﬂow. Therefore, we propose a function consisting of an exponentially decaying part and a constant part to describe the ﬂuctuations as follows sﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ rmsðpÞ (4.11) ¼ Dp eC p y=d mD þ Bp t0

Pressure- and velocity-measurements

(a)

101

(b)

(c) 3

Figure 4.11. Scaled power spectral densities PSD/u, roughly separated from effects of an oscillating water level. (a) t0/t0c ¼ 0.18 (run A02). (b) t0/t0c ¼ 0.36 (run A04). (c) with low mobility conditions t0/ t0c ¼ 0.59 (run A10). Derived from measuring time with Dt ¼ 120 s. Positions of the sensors see Fig. 4.10 and Table 4.2. The minimum level is deﬁned each by the detection limit of the MPPS and corresponds to an amplitude of about 0.1 Pa.

pﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ Figure 4.12. Damping of the pressure ﬂuctuations rmsðpÞ=t0 within the gravel layer. Run A01–A10. The pressure ﬂuctuations are ﬁltered with a high-pass f41.5 Hz.

M. Detert et al.

102

The best ﬁt for the ﬁltered pressure signal results in Dp ¼ 0.92, Cp ¼ 0.86 and Bp ¼ 0.57 with a correlation coefﬁcient of R2 ¼ 0.97. However, a more simple approach with Dp ¼ Cp ¼ 1 results in Bp ¼ 0.5 with R2 ¼ 0.92, what can also a good description of rms(p). Both ﬁts for equation (4.11) are plotted in Fig. 4.12. To compare the pressure ﬂuctuations to the pore ﬂow velocity, Fig. 4.13 shows vertical proﬁles of the local intrinsic 3D velocity U3D and their ﬂuctuations rms(u3D) (instead of their Cartesian components), as the ﬂow ﬁeld inside the pores is really 3D. (Typically, the u-component of the velocity trajectories is Z90% of the absolute velocity.) It becomes obvious that there is a strong scatter both in the velocity as well as in its ﬂuctuating part. This can be explained by measurement errors as follows: the gravel layer represents a highly stochastic geometric system of channels, indirections and dead-end pores. Depending on the ﬂow conditions and the current gravel geometry, in some experimental runs a rather homogeneous particle density was observed in the artiﬁcial pores, while in others this was not the case. As a result, the number of velocity vectors per frame is not constant. In extreme cases it may also drop to zero, i.e. there may be time periods where no velocity information is available at all since no tracer particles reached the observation volume. Two further effects are contributing to this ﬂuctuating information density. First, it was observed that during some experimental runs dirt particles were temporarily deposited in the artiﬁcial pores. Sometimes these dirt particles completely blocked one of the endoscopes’ view or reduced the intensity of the illumination and thus the signal-to-noise ratio. The second effect is related to the limitations of the image processing. Especially under low mobility conditions, the turbulence intensity in the upper grain layers becomes very large. In these cases, the maximum pore ﬂow velocities may reach values beyond the limits of the current endoscopic 3D-PTV. The interframe particle displacements in the image

(a)

(b)

Figure 4.13. 3D diminishing with gravel depth. Run A01–A10. P velocities within artiﬁcial pores, P (a) U 3D =u ¼ ð U i 2 Þ0:5 =u ; (b) rmsðu3D Þ=u ¼ ð rmsðui Þ2 Þ0:5 =u .

Pressure- and velocity-measurements

103

sequences become too large and cannot be tracked any more. Again, the number of recovered velocity vectors drops and the velocity statistics get biased towards lower velocities. In spite of these inaccuracies in the 3D-PTV technique using miniaturised endoscopic stereo setups, we try to give a functional relationship that is able to describe the pore velocity. In analogy to the derivation of the log law in open-channel ﬂow, we assume a logarithmic decrease for U3D in the transition from the main ﬂow above the gravel layer to the ﬂow within the uppermost gravel layers. Deeper, with increasing vertical gravel cover a constant seepage ﬂow becomes predominant. This results in U 3D ¼ DU ln ðC u y=d mD þ BU (4.12) u Due to the scatter of the data the best ﬁt results only in a correlation coefﬁcient of R2 ¼ 0.38 for DU ¼ 0.08, CU ¼ 0.51 and BU ¼ 0.19, respectively. Equation (4.12) is only valid for y/dmD ¼ 1.0, were U3D equals 1=4 u , and smaller velocities deeper in the gravel layer. For the velocity ﬂuctuations again an exponential function is assumed. It reads rmsðu3D Þ ¼ Du eC u y=d mD þ Bu u

(4.13)

The best ﬁt for the velocity ﬂuctuations results in Du ¼ 1.02, Cu ¼ 1.08 and Bu ¼ 0.07 with a correlation coefﬁcient of R2 ¼ 0.97. However, a simpliﬁcation with Du ¼ Cu ¼ 1 as for the declining rms(p) results in Bp ¼ 0.07 with the same R2 ¼ 0.75. Therefore, a simple, applicable scaling law for the damping of both, rms(p) and rms(u3D) within the upper porous gravel layer by ey=d mD becomes valid.

5.

Conclusions

Experimental results of pressure and velocity above and within a porous gravel layer are presented. New measuring techniques have been developed for these applications: MPPS measured turbulent pressure ﬂuctuations inside the gravel layer. Flow measurements were carried out by a 3D-PTV using miniaturised endoscopic stereo setups within artiﬁcial gravel pores. Additionally, in the free surface ﬂow a 1D-ADCP was used. Within the main ﬂow, alternating faster and slower ﬂuid packets with a size up to half of the water depth,oh/2, were observed by an 1D-ADCP. This agrees with the ﬂow pattern model of Adrian et al. (2000) for smooth beds. Unfortunately, the ADCP-measuring technique is not capable to resolve smaller vortices to zoom into details. Pressure ﬂuctuations rms(p) as well as velocity ﬂuctuations rms(u, v, w) decrease exponentially with increasing gravel depth, essentially within the ﬁrst two layers of gravel grains. Best-ﬁt equations that describe this trend were given. It is supposed, that the ﬂuctuations deeper down are dominated by a long wave oscillating mechanism. A rough ﬁlter to purge the signals from this effect was applied: within the frequency domain, the power spectral density (PSD) from the lowermost sensor

104

M. Detert et al.

within the gravel layer was subtracted from the other PSDs. The PSDs scale with the shear velocity u , independent if the hydromechanical load is corresponding to stable bed or to low mobility conditions.

References Adrian, R.J., Meinhart, C.D., Tomkins, C.D., 2000. Vortex organization in the outer region of the turbulent boundary layer. J. Fluid Mech. 422, 1–54. Coles, D., 1956. The law of the wake in the turbulent boundary layer. J. Fluid Mech. 1 (1), 191–226. Deﬁna, A., 1996. Transverse spacing of low-speed streaks in a channel ﬂow over a rough bed. In: Ashworth, P., Bennett, S., Best, J., and Mc Lelland, S. (Eds), Coherent Flow Structures in Open Channels. Wiley, England. Detert, M., Klar, M., Jehle, M., et al., 2004. Pressure ﬂuctuations on and in subsurface gravel layer bed caused by turbulent open-channel ﬂow. In: Greco, M., Carravetta, A., Della Morte, R. (Eds), River Flow 2004. Balkema. Dittrich, A., Nestmann, F., Ergenzinger, P., 1996. Ratio of lift and shear forces over rough surfaces. In: Ashworth, P., Bennett, S., Best, J., and Mc Lelland, S. (Eds), Coherent Flow Structures in Open Channels. Wiley, England. Emmerling, A., 1973. Die momentane Struktur des Wanddruckes einer turbulenten Grenzschichtstro¨mung. Mitteilungen aus dem Max-Planck-Institut fu¨er Stro¨mungsforschung. Farabee, T.M., Casarella, M.J., 1991. Spectral features of wall pressure ﬂuctuations beneath turbulent boundary layers. Phys. Fluid. 3 (10). Garcı´ a, M., Nino, Y., Lo´pez, F., 1996. Laboratory observations of particle entrainment into suspension by turbulent bursting. In: Ashworth, P.J., Bennett, S.L., Best, J.L., and Mc Lelland, S.J. (Eds), Coherent Flow Structures in Open Channels. Wiley, England. Gotoh, T., Rogallo, R.S., 1999. Intermittency and scaling of pressure at small scales in forced isotropic turbulence. J. Fluid Mech. 396, 257–285. Grass, A., 1971. Structural features of turbulent ﬂow over smooth and rough boundaries. J. Fluid Mech. 50, 233–255. Grass, A.J., Stuart, R.J., Mansour-Tehrani, M., 1991. Vortical structures and coherent motion in turbulent ﬂow over smooth and rough boundaries. Philos. Trans. R. Soc. Lond. Hoﬂand, B., 2002. Stability of coarse granular structures. Delft Cluster. Hoﬂand, B., 2004. Measuring the ﬂow structures that initiate stone movement. In: Greco, M., Carravetta, A., Della Morte, R. (Eds), River Flow 2004. Balkema. Hoﬂand, B., Battjes, J., 2006. Probability density function of instantaneous drag forces and shear stresses on a bed. J. Hydraul. Eng. 132 (11), 1169–1175. Hoﬂand, B., Battjes, J., Booij, R., 2005. Measurement of ﬂuctuating pressures on coarse bed material. J. Hydraul. Eng. 131 (9), 770–781. Klar, M., Stybalkowski, P., Spies, H., and Ja¨hne, B., 2002. A miniaturised 3-D particle-tracking velocimetry system to measure the pore ﬂow within a gravel layer. In: 11th International Symposium Applications of Laser Techniques to Fluid Mechanics. Lisbon, Portugal. Kline, S.J., Reynolds, W.C., Schraub, F.A., Runstadler, P.W., 1967. The structure of turbulent boundary layers. J. Fluid Mech. 30, 741–773. Nezu, I., Nakagawa, H., 1993. Turbulence in Open-Channel ﬂows. Monograph Series. Balkema. Roy, A.G., Bufﬁn-Be´langer, T., Lamarre, H., Kirkbride, A.D., 2004. Size, shape and dynamics of large scale turbulent ﬂow structures in a gravel bed river. J. Fluid Mech. 500, 1–27. Sechet, P., Guennec, L., 1999. Bursting phenomenon and incipient motion of solid particles in bed-load transport. J. Hydraul. Res. 37 (5), 683–696. Shields, A., 1936. Anwendung der A¨hnlichkeitsmechanik und der Turbulenzforschung auf die Geschiebebewegung. Mitteilungen der Versuchsanstalt fu¨r Wasserbau und Schiffbau, Berlin (87). Smith, C., Walker, J., Haidari, A., Sobrun, U., 1991. On the dynamics of near-wall turbulence. Philos. Trans. P. Scs. & Eng. (336), 131–175.

Pressure- and velocity-measurements

105

Vollmer, S., Francisco, d.l.S.R., Daebel, H., Ku¨hn, G., 2002. Micro scale exchange processes between surface and subsurface water. J. Hydrol. (269). Wilcock, P., Barta, A., Shea, C., et al., 1996. Observations of ﬂow and sediment entrainment on a large gravel bed river 32, 2897–2909. Willemetz, J., 1997. DOP1000. Signal Processing User’s Manual. WSA, 2004. Geschiebezugabe Iffezheim. Flyer. Zanke, U., 2001. Zum EinﬂuX der Turbulenz auf den Beginn der Sedimentbewegung. Mitteilungen des Instituts fu¨r Wasserbau und Wasserwirtschaft, Universita¨t Darmstadt (120).

Discussion by Genevie`ve A. Marquis and Andre´ G. Roy We compliment the authors for their very innovative approach to the problem. The experimental setup is meticulously designed and surely has the potential to answer some of the questions on the interactions, effects or feedbacks between ﬂows above and within a porous gravel bed. It addresses the very difﬁcult technical problems of measuring interstitial ﬂow. The data presented be the authors raise several interesting issues. Here, we would like to discuss the importance of large-scale turbulent ﬂow structures and the interaction between interstitial and above the surface ﬂows. Firstly, the signiﬁcance of the 1D-ADCP results may have been underestimated. The slower and faster packets of ﬂuid detected and shown in Fig. 4.6 are similar to largescale turbulent structures found in ﬂows above rough boundaries. Such structures scale with ﬂow depth (Roy et al., 2004). The origin of these large-scale ﬂow structures is not well known and interstitial ﬂow may play an active role in their generation and maintenance. Looking at Fig. 4.10, it appears that there is a peak in the spectra of all pressure sensors (except the one above the bed) between 0.4 and 0.7 Hz. This range may indicate a direct relationship with large-scale turbulent structures. Second, the data showing the mechanisms of interactions between interstitial and surface ﬂows raise interesting questions. In Figs. 4.8 and 4.12, amplitude and frequency of pressure ﬂuctuations are diminishing with depth, the higher values being associated with the surface ﬂow. This observation raises the bottom-up or top-down question. Are the pressure ﬂuctuations in the interstitial ﬂow inducted by the surface ﬂow and is it the other way around? What is the nature of the feedback effects between the interstitial ﬂow structure and the surface ﬂow structure and at what scale do they operate? The shape of the pressure distributions may contain part of the answer. In Fig. 4.9, we see that the pressure distributions change from a symmetric shape within the bed to a positively skewed one above the bed. Is the mechanism responsible for this change in the shape of the distributions related to dampening of the impacts of largescale downward sweeping motions on the interstitial ﬂow? Could the authors speculate on the role, generation or maintenance of large-scale turbulent ﬂow structures of the surface ﬂow in relation to the interstitial ﬂow? References Roy, A.G., Bufﬁn-Be´langer, T., Lamarre, H., Kirkbridge, A.D., 2004. Size, shape and dynamics of large scale turbulent ﬂow structures in a gravel bed river. J. Fluid Mech. 500, 1–27.

Pressure- and velocity-measurements

105

Vollmer, S., Francisco, d.l.S.R., Daebel, H., Ku¨hn, G., 2002. Micro scale exchange processes between surface and subsurface water. J. Hydrol. (269). Wilcock, P., Barta, A., Shea, C., et al., 1996. Observations of ﬂow and sediment entrainment on a large gravel bed river 32, 2897–2909. Willemetz, J., 1997. DOP1000. Signal Processing User’s Manual. WSA, 2004. Geschiebezugabe Iffezheim. Flyer. Zanke, U., 2001. Zum EinﬂuX der Turbulenz auf den Beginn der Sedimentbewegung. Mitteilungen des Instituts fu¨r Wasserbau und Wasserwirtschaft, Universita¨t Darmstadt (120).

Discussion by Genevie`ve A. Marquis and Andre´ G. Roy We compliment the authors for their very innovative approach to the problem. The experimental setup is meticulously designed and surely has the potential to answer some of the questions on the interactions, effects or feedbacks between ﬂows above and within a porous gravel bed. It addresses the very difﬁcult technical problems of measuring interstitial ﬂow. The data presented be the authors raise several interesting issues. Here, we would like to discuss the importance of large-scale turbulent ﬂow structures and the interaction between interstitial and above the surface ﬂows. Firstly, the signiﬁcance of the 1D-ADCP results may have been underestimated. The slower and faster packets of ﬂuid detected and shown in Fig. 4.6 are similar to largescale turbulent structures found in ﬂows above rough boundaries. Such structures scale with ﬂow depth (Roy et al., 2004). The origin of these large-scale ﬂow structures is not well known and interstitial ﬂow may play an active role in their generation and maintenance. Looking at Fig. 4.10, it appears that there is a peak in the spectra of all pressure sensors (except the one above the bed) between 0.4 and 0.7 Hz. This range may indicate a direct relationship with large-scale turbulent structures. Second, the data showing the mechanisms of interactions between interstitial and surface ﬂows raise interesting questions. In Figs. 4.8 and 4.12, amplitude and frequency of pressure ﬂuctuations are diminishing with depth, the higher values being associated with the surface ﬂow. This observation raises the bottom-up or top-down question. Are the pressure ﬂuctuations in the interstitial ﬂow inducted by the surface ﬂow and is it the other way around? What is the nature of the feedback effects between the interstitial ﬂow structure and the surface ﬂow structure and at what scale do they operate? The shape of the pressure distributions may contain part of the answer. In Fig. 4.9, we see that the pressure distributions change from a symmetric shape within the bed to a positively skewed one above the bed. Is the mechanism responsible for this change in the shape of the distributions related to dampening of the impacts of largescale downward sweeping motions on the interstitial ﬂow? Could the authors speculate on the role, generation or maintenance of large-scale turbulent ﬂow structures of the surface ﬂow in relation to the interstitial ﬂow? References Roy, A.G., Bufﬁn-Be´langer, T., Lamarre, H., Kirkbridge, A.D., 2004. Size, shape and dynamics of large scale turbulent ﬂow structures in a gravel bed river. J. Fluid Mech. 500, 1–27.

M. Detert et al.

106 Reply by the authors

y [mn]

y [mn]

The discussion authors state that the signiﬁcance of the ADCP results may have been underestimated. After visualising the data in a different manner, we must admit: Yes! Fig. 4.14 gives an alternative view of the identical signal shown already in Fig. 4.6. Assuming frozen turbulence, structures that pass the ultrasonic beam can be detected as regions of high or low velocities, respectively. The time-dependent signal can be shifted to a length scale simply by multiplication with a typical transport velocity. Here, we use the bulk velocity U. The result is a streamwise length scale lx ¼ Ut instead of the time scale t. In this way, the wedge-like large-scale structures that incline in streamwise direction become more obvious. And indeed, in this example they scale with the water depth h, roughly by a factor of 0.5–4. Moreover, the discussion authors speculate that there might be a direct relationship to the peak in the pressure spectra. And again, it seems to be right: For this example, the peak value in the pressure spectra was determined to 0.5 Hz (not shown in here). Therefore, the number of the wedge-like structures within a length of time of 7.0 s must be in an order 3–4. Keeping in mind, that a time series of 7 s typically is not enough for a statistically founded statement, this ﬁts very well with the number ﬁve fast ﬂuid packets (colour-coded by light grey and white, respectively) we can ﬁnd within the present example. Thus, the peak frequency is not only due to long wave water oscillations, but also directly related to – and induced by – the large-scale wedge-like structures. Unfortunately, the deﬁnite relation between x and lx is unclear, as a Cartesian decomposition of ux into u and v is impossible. A multidimensional measuring technique like the PIV-technique would be more adequate to study these structures in the open channel ﬂow. The second topic in the discussion concerns the question, if the interstitial ﬂow is mainly dominated by the surface ﬂow or if the mechanism runs vice versa. Fig. 4.13

200 100 0 2500

200 100 0 5000

2.

2000

1500 1000 λx=U.t [mm], t=-0.2-3.3[s] 5.

4500

1.

500

0

4.

4000 3500 λx=U.t [mm], t=3.1-6.6[s] -100

3.

3000 0

2500 100

Figure 4.14. Coherent turbulent structures passing the measuring area within 7.0 s. Footprint of velocity ﬂuctuations show ux 0 ¼ f(lx, y), measured by ADCP. lx ¼ 301–565 mm is directly according to Fig. 4.6. Flow direction from left to right.

Pressure- and velocity-measurements

107

shows that the pore velocity at y/dmD ¼ 1 is U3D ¼ 0.5u*. In contrast to this, the bulk velocity in the ﬂow over a rough bed scales typically with U ¼ 10u*. This means that most of the kinetic energy is transported in the mean ﬂow, whereas the interstitial ﬂow rather has a passive role. We think that the driving mechanisms are due to the surface ﬂow, especially from the shear. The ﬂow in the gravel bed only reﬂects the turbulent ﬂow conditions in the main ﬂow.

Gravel-Bed Rivers VI: From Process Understanding to River Restoration H. Habersack, H. Pie´gay, M. Rinaldi, Editors r 2008 Elsevier B.V. All rights reserved.

109

5 Evaluating vertical velocities between the stream and the hyporheic zone from temperature data Ina Seydell, Ben E. Wawra and Ulrich C.E. Zanke

Abstract The exchange between stream and interstitial pore water is vital for the ecosystem of the hyporheic zone, since it determines oxygen supply and thermal conditions. Determining ﬂow velocities within the sediment is extremely difﬁcult. Sampling of temperature data in the riverbed can be used as a tracer to determine vertical ﬂow velocities. Three different approaches to determine velocities are compared in this study. Velocities calculated using time-lag (Constanz and Thomas, 1996; Ingendahl, 1999) and temperature damping (Taniguchi, 1993) are compared with a 1D-numerical model. Simulated temperatures derived from the model are then evaluated with measured data from a ﬁeld-site at the Lahn near Marburg. Results from the time-lag method differ signiﬁcantly from simulated velocities. Correlation was good for modelled vertical velocities above 3 cm h1. Values calculated using temperature damping showed an excellent linear correlation with the model up to a vertical velocity of 6.6 cm h1. Temperatures determined with the 1D-numerical model are in good accordance with measured data. Zones with down-welling as well as zones of predominantly upwelling conditions could be veriﬁed as Darcy’s velocity ranged between 1 and +5 cm h1. Although velocity calculated from time-lag or temperature damping may be sufﬁcient for some situations, these methods are restricted to situations with signiﬁcant temperature amplitudes. The 1D-approach is appropriate to determine local scale (vertical) hyporheic exchange. 1.

Introduction

The hyporheic zone is the spawning habitat and refuge for a variety of species such as macroinvertebrates. It shows highest colonisation densities within the uppermost E-mail address: [email protected] (I. Seydell) ISSN: 0928-2025

DOI: 10.1016/S0928-2025(07)11122-6

110

I. Seydell, B.E. Wawra, U.C.E. Zanke

20–40 cm of the riverbed and functions as a transition zone between the stream, the groundwater, and the semi-terrestrial ﬂood plain with strong temporal and spatial gradients of water characteristics and high biological variability (Gibert et al., 1990; Vervier et al., 1992). Exchanges of water and solutes between the stream and its bed are vital for the ecosystem (Chapman, 1988; Ingendahl, 2001; Malcom et al., 2004). The processes were studied from the biological viewpoint since the 1960s, and from the hydrological point of view since the 1980s. In particular, the oxygen concentration depends on ﬂow velocity and ﬂow-path length within the sediment (Hakenkamp and Palmer, 2000). Thus, the exchange velocity and direction is of major importance. Spawning sites, for example, correlate with ﬂow velocities within the sediment (Geist, 2000; Malcom et al., 2004). Furthermore, the mass exchange of water and solutes is not only important for hyporheic organisms, but for the groundwater fauna and the self-puriﬁcation of the stream as well (Brunke, 2000). Flow-patterns in the riverbed are complex and difﬁcult to determine. Depending on morphology and subsurface hydraulic conditions, the proportion of lateral to vertical velocity varies. As a function of this proportion, Saenger and Zanke (in print) developed a conceptual model with three hyporheic layers of exchange: active exchange, moderate exchange, and groundwater domination. The distinction between these three layers is based on the ratio between the horizontal and the vertical ﬂow component. This model underlines the importance of the vertical exchange for the supply of hyporheic pore spaces with oxygen and nutrients. Therefore, the vertical velocity component is the fundamental parameter characterising the hyporheic exchange. In the past, a variety of methods was used to evaluate ﬂow velocities in the riverbed, such as tracer experiments with or without the extraction of pore water (Harvey et al., 1996; Benson, 1999), dilution or erosion of material inserted into the bed (Petticrew and Kalff, 1991; Thompson and Glenn, 1994; Carling and Boole, 1996), or piezometer measurements (Saenger, 2000; Storey et al., 2003). All these methods are either time consuming and laborious, and therefore only good for short-term measurements, and may be prone to mistakes. Considering the differences between the groundwater and the stream it appears that temperature or temperature differences may be used as a natural tracer. The advantage is that continuous data collection is possible over long periods, thus providing detailed information about velocities without changing hydraulic conditions. Water temperatures in a stream show annual as well as diurnal patterns (Taniguchi, 1993; Lenk, 2000). These temperature curves continue within the riverbed and are damped with depth (Saenger, 2000). From the change of the temperature pattern with depth, down- and up-welling can be identiﬁed (Silliman and Booth, 1993; Constanz and Thomas, 1996). This can be done either by looking at the degree of damping with depth or the time-lag between temperature curves from different depths (Fig. 5.1). From this premise the following questions arise: How do we need to analyse collected temperature data to evaluate the exchange between the stream and the hyporheic pore space? Can we obtain reliable values for the vertical ﬂow velocity?

Evaluating vertical velocities

111 dtmax

17.6

temperature (°C)

17.2 ∆Tz 16.8 ∆Tz0 16.4

dtmin

16 0

12

24

36

48

hours Figure 5.1. Example of time-lag and difference of amplitudes for temperature-damping method between two temperature curves.

Is it possible to show the correlation between rifﬂe structure and subsurface ﬂow with temperature data?

2.

Experimental setup and data analysis

Investigations were conducted at a rifﬂe structure of the Lahn, near Marburg, Germany. Characteristics are given in Table 5.1. On the left hand side of the stream was a pond in which was enlarged towards the stream in September 2000. Mean surface water temperature during the study in the stream was about 12.31C. Minimal stream temperature was 1.11C and maximum was 25.81C. A maximum diurnal temperature difference of 7.41C was measured in June 2000. Mean air temperature was 10.41C. Temperatures varied between 10.8 (January 2001) and 26.31C (July 1999 and August 2001). Temperature was measured in the stream centre at two locations along the rifﬂe (Fig. 5.2). Location C-1 was located in a down-welling zone at the upstream side of the rifﬂe crest and location C-2 at the rifﬂe tail. Distance between C-1 and C-2 was 49 m. We installed PRT-temperature sensors with an accuracy of 0.11C (pt-100) at the sediment surface and at depths of 20, 50, 100, and 150 cm within the sediment and continuously measured the temperature between 1999 and 2001. Data was averaged and stored every 30 min. A number of methods were utilized to determine one dimensional ﬂow velocity from the temperature data. In the 1960s, techniques have been established to estimate

I. Seydell, B.E. Wawra, U.C.E. Zanke

112 Table 5.1.

Characteristics of the study reach at the river Lahn.

HQ1

79

m3 s1

Slope Channel width Rifﬂe height d50 of surface layer d50 of subsurface material Sorting index Absolute porosity (determined from density) Estimated maximal effective porosity (from grain size distribution) Organic matter content for material smaller 2 mm Mean diurnal air temperature Stream water temperature

0.0015 18 0.71 70 21 4.5 25 20

– m m mm mm – % %

2 10.4 (10.8 to 26.3) 12.3 (1.1 to 25.8)

% 1C 1C

C-1

C-2

data logger

data logger

Interstiti

al pt-100 sensors

pt-100 sensors

stream

Figure 5.2. Temperature measurements with pt-100 sensors at the sediment surface and at 20, 50, 100, and 150 cm depth.

one-dimensional Darcy’s velocities from temperature curves. Cartwright (1970), calculated velocities within the sediment from continuously recorded temperature data with known thermal material properties of the sediment. The non-isothermal, one-dimensional ﬂow within a homogenous subsurface water body was expressed by Stallmann (1965) as CM

@T @2 T @T ¼ kM 2 qC W @t @z @z

(5.1)

where CW and CM are the volumetric heat capacity of water and of the water– sediment mixture in J m3 K1, T the temperature in 1C, t the time in h, z the depth in m, kM the thermal conductivity of the water–sediment mix in W m1 K1, and q Darcy’s velocity in cm h1.

Evaluating vertical velocities

113

The volumetric heat capacity of the water–sediment mixture is calculated as C M ¼ cs rs n þ cw rw ð1 nÞ

(5.2) 1

1

Here cS and cW are the speciﬁc heat of sediment and water in J kg K , rS and rW the density of sediment and water in kg m3, and n the total pore volume. Nowadays, numerical models of varying complexity (Storey et al., 2003) are used along with simple methods to estimate one-dimensional ﬂow from temperature data (Pusch, 1993; Lenk, 2000). The heat capacity of the sediment–water mixture is calculated from the heat capacity of sediment and water. In this study, a 1D-model is utilized to calculate temperatures with vertical velocity components determined inversely. Two commonly used methods to calculate ﬂow velocities from temperature data taken at different depths are tested and compared with results of the 1D numerical model. The temperature travel-time method uses the time-lag between two temperature curves to determine the travel-time of temperatures (vT). The temperature-damping approach uses damping of temperature amplitudes with depth. Both methods as well as the numerical model are based on the assumption of one-dimensional, non-isothermal ﬂow. 2.1. 2.1.1.

Method description 1D model for comparison of methods

We applied the numerical code HydroBioGeoChem123d (hbgc123d) to build a numerical one-dimensional ﬁnite element model (Gwo et al., 1999). This code is capable of simulating coupled non-isothermal hydrologic transport of solutes and temperatures in variably saturated media. The implemented equation for heat transfer in conservative form is @T @r Y þ rðrW C W TvÞ þ r C W W T ðrW C W Y þ rbS C bS Þ @t @t rðDT rTÞ ¼ rW C W qT

ð5:3Þ

where rW is the ﬂuid density (kg m3); Y effective moisture content (m3 m3) [ ¼ neS in which ne is the effective porosity (m3 m3), and S the degree of effective saturation of water (m3 m3)]; rbS the bulk density of the dry medium (kg m3); CW and CbS the speciﬁc heats of the groundwater and the dry medium in subsurface systems, respectively (m2 h2 K1); T the temperature (K); v the Darcy’s velocity of the groundwater(m h1); DT the thermal dispersion/diffusion/conductivity coefﬁcient tensor (kg m h3 K1). On the right side is the source/sink term of heat that may be due to artiﬁcial injection (kg m1 h3). Transport and heat transfer equations are solved iteratively with a Lagrangian–Eulerian ﬁnite element method. We modelled the transient heat transfer with a domain representing a column (1D) of 1.5 m depth. The model consists of 751 nodes and 750 elements. Input data are the spatial distribution of nodes and elements, as well as thermal properties and moisture content of the media (Table 5.2). Boundary conditions are measured temperatures at the surface and at 1.5 m depth. The chosen velocities ranged between minus and plus

I. Seydell, B.E. Wawra, U.C.E. Zanke

114 Table 5.2.

Chosen parameters for 1D-model.

Parameter

Value

Units

rS ¼ density of sediment cS ¼ speciﬁc heat of sediment k ¼ heat conductivity of sediment Y ¼ moisture content ¼ n

2700 2.0 2.5 25

kg m3 J cm3 1C1 J s1 m1 1C1 %

20 cm h1 and are constant during each run. Output data are temporal and spatial temperature distributions. We could not determine changes in hydraulic conductivity from sediment analysis for the study reach. This means layering of the sediment could not be identiﬁed for the ﬁeld site. Although anisotropy and inhomogenities have signiﬁcant inﬂuence on subsurface ﬂow pattern, their inﬂuence is visible only from velocities in the 1D-model. Therefore the sediment was simpliﬁed to be homogenous. 2.1.2.

Estimating velocities from temperature travel-time

This method assumes that the governing transport mechanism for temperature is advection while material properties like heat capacity and heat conduction may be neglected. Temperature travel-time is equal to the time-lag between two maxima or minima (dt in h) between two temperature curves from different depth (see Fig. 5.1). Travel-time divided by the travel distance (dz in m) the temperature velocity is calculated as dz (5.4) dt For this travel-time approach, the one common way to determine the time-lag found in literature is statistical cross correlation (Ingendahl, 1999; Lenk, 2000). In order to exclude trends in the dataset, temperature differences between two time steps are analysed instead of the measured temperatures. This procedure does not ensure that the lag between two minima or two maxima is determined and may produce erroneous time-lag, and thus not reliable for our dataset. To solve this problem the travel-time of temperatures was determined visually for small datasets and by a MATLAB-routine MOB (Seydell, in print) for large datasets. This program calculates the time-lag between the minima and the maxima of two temperature curves, giving the mean of those two values as time-lag. If the mean of the time-lags from minima and maxima is used, the error resulting from trends in the dataset is minimized. The travel-time analysis is limited to periods with large temperature differences between day and night as well as recognisable temperature amplitudes at all depths. Furthermore, the calculated velocity of temperature needs to be transferred into Darcy’s velocity. We compare two different approaches for this procedure. The ﬁrst approach neglects material properties. Temperature is utilized as an ideal tracer without adsorption. In this case, the travel-time of temperature is equal to the vT ¼

Evaluating vertical velocities

115

travel-time of water and Darcy’s velocity (m h1) is calculated as q ¼ vT ne

(5.5)

where vT is the temperature travel-time (h) and ne the effective porosity (). If heat capacity of water and the water sediment mixture are considered, but conductive transport is neglected as suggested by Constanz and Thomas (1996), equation (5.1) becomes q ¼ vT

CM CW

(5.6)

In this study we evaluate the error resulting from neglecting heat capacity conduction and heat capacity by comparison with the one-dimensional model. 2.1.3.

Estimating velocities from temperature damping with depth

The approach to determine velocity from temperature damping considers heat transport by advection and conduction, thus comprising relevant material properties. Based on the work of Stallmann (1965) and Cartwright (1971), Taniguchi (1993) developed a method to determine horizontal groundwater ﬂow velocity. This method assumes one-dimensional ﬂow and operates effectively only with sinusoidal temperature curves with pronounced amplitudes at the sediment surface. Thermal properties of the water sediment mixture are input data. Taniguchi (1993) established a dimensionless parameter b, which comprises velocity and material properties. This parameter can be obtained by ﬁtting measured data to type curves and ranges between 2 and 2. These type curves are deﬁned by ln(DTz/DT z0 ) versus (CM p k1 t1)0.5 (zz0) ¼ K0.5(zz0), where z0 is depth 0 at the sediment surface, z depth (m), DTz and DT z0 are the temperature amplitudes at depths z and z0 (1C), CM the speciﬁc heat (J m3 s1) of the mix, r the density of the mixture (kg m3), t the period of the sinusoidal temperature (s), and kM the thermal conductivity of the mix (W m1 1C1). Flow velocity (m s1) is calculated as pﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ ððkM C M pÞ=tÞ q¼b CW 2.1.4.

(5.7)

Comparison of the methods

For testing and performance comparison of the travel-time and the damping methods, temperatures from a 3-day period in August 2000 were selected as boundary conditions for the numerical model. We used the output temperature curves from the model as input data for the two other methods. Velocities varied from q ¼ 20 (up-welling) to 20 cm h1 (down-welling). Finally, we compared the chosen input velocities for the model with the output velocities from the travel time and temperature-damping analysis.

I. Seydell, B.E. Wawra, U.C.E. Zanke

116 Table 5.3.

Chosen periods for simulation of typical situations at the ﬁeld site.

Situation

Period

Discharge (m3 h1)

Vertical velocity for depth between 0–20

Early summer A Summer A Summer P Winter P

2.2.

20.06.–25.06.2000

0.27

26.08.–31.08.2000

1.0 to 2.0

28.09.–04.10.2000

2.5 to 4.0

01.11.–06.11.2000

1.0

C-1 C-2 C-1 C-2 C-1 C-2 C-1 C-2

+4.0 –1.0 +3.0 –1.0 0 –1.0 +1.0 –1.5

20–50 –1.5/ –1.0 0 0 0 –2.0 +0.3 –1.0

50–100 –1.5 –1.0 –1.0 0 –3.0 +1.0 +0.2 +5.0

1D model for selected periods

For the ﬁeld site, four typical periods were selected to model temperatures and to determine vertical velocity components up to 1 m depth (Table 5.3). All four periods are characterized by a constant low discharge in the stream. The September period is extraordinary as a drastic change of subsurface ﬂow direction occurred due to the ﬁlling of a nearby pond. Since we expected ﬂow patterns to vary with depth, the onedimensional model was divided into three sub-domains to account for the change in ﬂow conditions with depth. Each sub-domain represents a layer with three points of temperature measurement. For boundary conditions at the upper and lower end of the sub-domain we utilize measured data, whereas the temperatures in the middle of each sub-domain were modelled and then compared to the measured values. We deﬁned the sub-domains according to the depth of the modelled temperatures within the sediment. The upper domain (sub-20) extends from 0 to 50 cm and temperatures at 20 cm were simulated, the middle domain (sub50) extended from 20 to 100 cm, and the lower domain (sub-100) from 50 to 150 cm depth. Material properties were identical to those described above and velocities were constant for each sub-domain and run. 2.3.

Quantifying model performance

To verify the model performance, a method to quantify the accuracy or the deviance between the modelled and the measured temperatures was needed. The relative performance of one model-run to another was controlled by calculating a variable ABW similar to the standard deviation, giving the mean error in 1C. sﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ P ðxmodelled xmeasured Þ2 (5.8) ABW ¼ n where xmodelled is the simulated temperature, xmeasured the measured temperature, and n the number of data points. To make this result more comparable between two series, a parameter that incorporates the range of values is needed. Since the procedure established by Nash and Sutcliffe (1970) malfunctions with periodic,

Evaluating vertical velocities

117

respectively sinusoidal datasets a new indicator (ABWR) was calculated by dividing ABW by the range of all measured temperatures during a modelled period. ABWR ¼

ABW 100 Range of measured values

(5.9)

ABWR is the error in per cent of the range from all measured temperatures between the upper and the lower boundary.

3. 3.1.

Results Performance of the 1D-model

Behr (2002) and Kirchmaier (2003) have tested the sensitivity of vertical ﬂow velocity to parameters characterising the sediment for ranges given in literature (Rauch, 1991; Taniguchi, 1993; Scheffer and Schachtschabel, 2002). Porosity has been varied between 0.1 and 0.3, heat capacity of the sediment between values from 1.5 to 2.5 (J m3 1C1), and heat conductivity of the saturated aquifer between 1.5 and 3.0 (W m1 1C1). Our results show that variation of these parameters show little inﬂuence, compared to the inﬂuence of even small velocities in the model. Our chosen model parameters for comparison with other methods as well as for modelling selected periods are given in Table 5.2. For the 3-day test period, modelled temperature curves show good accordance with measured data. Up to 20 cm depth, the mean deviation between modelled and measured data is about 0.31C or 3% of the measured range (Fig. 5.3a). Below 20 cm depth, the deviation decreases to about 0.11C respectively 1% of the measured range, which is in the order of the accuracy of the temperature sensors. In order to relate the mean deviance of temperature to the vertical velocity component we evaluate the relation between temperature curves and simulated velocity. The simulated velocity was varied in steps of 1 cm h1 between 10 and +20 cm h1. The change in mean temperature deviance between two subsequent model runs (ABW(runX) ABW(runX1)) shows highest variance between 4 and +4 cm h1 (see Fig. 5.3b). Outside this range, signiﬁcance of ABW decreases drastically as changes of ABW with changes in velocity become very small. The relation between changes in temperature (ABW) of two model runs and changes in velocity is not a linear function. A rating curve is needed to translate the performance of temperature modelling into accuracy of velocity determination. For the investigated case we distinguished velocities between 5 and +10 cm h1 (Fig. 5.3b). Within this range the relation of ABW to velocity, respectively, the change in ABW between two subsequent model runs allows evaluation of the error in the calculated velocity. To evaluate the error induced when neglecting heat conduction in the travel-time method, we compared model runs with and without heat conduction for velocities between –10 and +20 cm h1 (see Fig. 5.3c). The results show a maximum error for small positive (downward) velocities. At 20 cm depth the maximum error occurs at 1 cm h1 with ABW ¼ 0.71C, for greater depths the maximum error is higher and occurs for larger velocities.

I. Seydell, B.E. Wawra, U.C.E. Zanke

118

temperature (˚C)

17 a

16 15

measured simulated analytical

14 13 12 48

72

96

120

time (h) b

c

ABW (˚C)

1.2

0.8 20 cm 50 cm 100 cm

0.4

0 -10

0

10

-10 q (cm h-1)

0

10

20

Figure 5.3. (a) Measured, analytically calculated, and simulated temperatures at 20 cm depth for the 3-day test period. (b) Mean deviation (ABW) between simulated temperature curves for given velocities and velocity ¼ 0 cm h1. (c) Mean deviation (ABW) between simulated temperatures with and without consideration of heat conductivity k.

3.2. 3.2.1.

Comparison of methods Travel-time

The calculation of travel-time showed some obviously wrong values for the crosscorrelation approach. These errors can be excluded by visually checking the temperature curves, but then the advantage of automation is lost. Therefore, we rated this procedure as not reliable. Instead, we calculated travel-time of temperatures with the module MOB, which excludes obviously wrong time-lags, for example, negative ones, from further analysis. Due to the nature of the travel-time method, where the time-lag between temperature proﬁles may not be negative, up-welling (negative velocities) could not be determined. Since amplitudes were not apparent below 50 cm sediment depths, velocities were not calculated. The vertical ﬂow velocity calculated with temperature travel-time varies, depending on the conversion of temperature travel-time into ﬂow velocity of water (Fig. 5.4). When utilizing the travel-time of temperatures the pore velocity ranges between 0.6 and 10 cm h1. Calculated velocities were much smaller than simulated

Evaluating vertical velocities

119

time-lag (n)

time-lag (CM / CW) 20 v (m h-1) simulated

20 v (cm h-1) simulated

damping

15 10 5 0

15 10 5 0

20 cm depth

50 cm depth -5

-5 -5

0

5

10

15

v (cm h-1) calculated

20

-5

0 5 10 15 20 v (cm h-1) calculated

Figure 5.4. Simulated velocities compared to calculated velocities from travel-time and temperature damping.

values; the error increased with velocity from 41% for a model velocity of 3 cm h1 to 66% for 20 cm h1. Including the heat capacity of water and sediment, led to calculated velocities ranged between 11 and 25 cm h1. Calculated velocities were higher than simulated values. Deviance decreased from 41% for the model velocity of 3 cm h1 to 9% for the model velocity of 20 cm h1. Both methods show that the determined velocities correlate well with simulated values as long as they were larger than +2 cm h1. 3.2.2.

Temperature damping

Velocities calculated with temperature damping ranged between 2 and +6.6 cm h1 and showed a good correlation with the numerical model (Fig. 5.4). However, the parameter b is limited to the range between 2 and +2, leading to a maximum velocity of 76.6 cm h1 for the investigated material properties and diurnal temperature amplitudes. All model velocities larger than this were calculated to be 6.6 cm h1. For velocities smaller than 2 cm h1 (up-welling) the temperature amplitudes within the sediment were rapidly damped, thus the method failed to give any result even at a depth of only 20 cm. As with the travel-time procedure, it was not possible to calculate velocities for depths greater than 50 cm, since diurnal amplitudes were too small in this area. Between model velocity of 2 and +6.6 cm h1 the values calculated from temperature damping are very close to the modelled ones, deviance varying between minus 33% for model velocities close to zero and –17% for a model velocity of 6 cm h1. 3.3.

Typical temperature regimes for selected periods at the study site

The simulation of water temperature for the selected period of 6 days each produced results that closely resembled the measured data. The period in June 2000 represents a typical early summer situation. The mean temperature of the surface water decreased from 21 to 151C with pronounced diurnal temperature amplitudes during

120

I. Seydell, B.E. Wawra, U.C.E. Zanke

the 6 day period. At location C-1 a strong down-welling with a vertical velocity component of +5 cm h1 within the uppermost 20 cm of the sediment was accompanied by a small time-lag between the falling limbs of temperature curves from the model to the measured ones (ABWR ¼ 4%). Below 20 cm depth the vertical velocity was determined to be 1.5 cm h1 and modelled temperatures closely met the measured ones (ABWR ¼ 2%). At the rifﬂe tail (location C-2) up-welling of about 1.0 cm h1 was constant for all depths down to 1 m. For the 20 cm depth the modelled temperatures showed the largest mean deviation ABWR ¼ 4% to measured values of all modelled periods. The amplitudes within the ﬁrst 2 days were much smaller in the model than measured in the ﬁeld. The period in August 2000 shows a typical summer situation with a mean surface water temperature of 16.41C. Daily amplitudes ranging between 15.5 and 18.51C at the sediment surface show a slightly different pattern. Again at location C-1, a downwelling of about +3 cm h1 within the top 20 cm of sediment was accompanied by the familiar time-lag between modelled and measured temperature curves. No vertical ﬂow was determined at 50 cm depth and a slight up-welling of about +1 cm h1 at 100 cm depth (ABWR ¼ 1%). At location C-2 an up-welling of 1 cm h1 was visible at 20 cm depth with no visible vertical ﬂow at 50 and 100 cm within the sediment (ABWRo1%). For the periods in June and August 2000 before the enlargement of the pond, the deviation between measured and modelled temperature curves was highest at 20 cm depth. In the down-welling zone at location C-1 with pronounced amplitudes, the modelled temperatures showed a time-lag in the falling limb of the measured temperature curves between 1 and 2 h for the period during summer (Fig. 5.5, June and August). In the predominantly up-welling zone at location C-2, modelled temperature curves for the summer situations (June and August) at 20 cm depth generally showed smaller amplitudes than the measured ones. For the period in June, the simulated temperature curve meets the measured data at the daily minima. For the period in August the two curves meet at the maxima of the daily pattern. The situation in September 2000 was uncommon for the ﬁeld site, as a nearby pond of the sewer treatment plant was enlarged towards the river and ﬁlled shortly before our measuring campaign. At the ﬁeld site, we saw groundwater ﬂowing through the bank into the river. At location C-1, measured surface water temperatures dropped during the last 2 days. Temperatures at 50 cm depth did not follow this trend within the given period (Fig. 5.5), thus remaining above the temperatures recorded at 20 and at 100 cm depth. From the model there is no vertical velocity apparent at 20 and 50 cm depth (ABWR ¼ 2%), however an up-welling current with a velocity of 3 cm h1 (ABWR ¼ 1%) was visible at 100 cm depth. Note that temperatures at 100 cm depth were just below those at 150 cm depth and thus just outside the expected range. In contrast, location C-2 showed up-welling between 1 cm h1 (ABWR ¼ 3%) in 20 cm depth and 2 cm h1 (ABWR ¼ 3%) in 50 cm depth and down-welling of +1 cm h1 (ABWR ¼ 1%) at 100 cm depth. For the winter scenario in November 2000, the modelled temperature related to a pattern of mainly down-welling at location C-1. Velocities ranged from +1 cm h1 at 20 cm depth to very small velocities of +0.3 cm h1 at 50 cm and +0.2 cm h1 at

Evaluating vertical velocities

121

20 cm mod. 20

50 cm mod. 50

100 cm mod. 100

24 22 20 18 16

C-1 June 2000

C-2 June2000

C-1 August 2000

C-2 August 2000

C-1 September 2000

C-2 September 2000

C-1 November 2000

C-2 November 2000

14 18 17 temperature (˚C)

16 15 14

14 13 12 11 10 9 8 7 0

24

48

72 96 120 144 hours

0

24

48

72 96 120 144 hours

Figure 5.5. Measured and simulated temperatures at depths of 20, 50, and 100 cm within the river bed for selected periods.

100 cm depth with a deviance ABWR of less than 1% for all depths. Up-welling was apparent at location C-2 at 20 and 50 cm depth with velocities of 1.5 and 1 cm h1 respectively. At a depth of 100 cm at location C-2 the model showed a strong down-welling velocity component of 5 cm h1. Again, the deviation between modelled and measured values ABWR was less than 1% for all depths. Note that measured temperatures in 50 cm and 100 cm depth were almost identical in this case. Measurements of hydraulic heads done shortly before and after the modelled period indicated a strong lateral gradient near location C-2 at the end of October, which decreased signiﬁcantly towards the end of November.

I. Seydell, B.E. Wawra, U.C.E. Zanke

122 4.

Discussion

4.1. 4.1.1.

Method rating Travel-time

The major disadvantage of this method is that the time-lag is always positive, thus negative velocities respectively up-welling cannot be identiﬁed. For model velocities above +2 cm h1 correlation between the model and calculated values was well (Fig. 5.4). When we ignored material properties the total error in the calculated values for data from the ﬁeld site was larger than 50% of the calculated value (Table 5.4). A correction function of the form y ¼ a x+b was obtained for comparison with the model but it varied for different depths within the sediment. Taking heat capacity of sediment and water into account, the error was 26% at a model velocity of +4 cm h1, decreasing to 9% at a model velocity of 20 cm h1. This pattern of large relative errors for low positive velocities relies on the error resulting from neglecting heat conduction, which showed maximal errors between 0 and +5 cm h1 (see Fig. 5.3c). For a loosing reach of the Rio Grande, Constanz and Thomas (1996) found that ﬂow calculated from travel-time with cm/CW was twice as large as ﬂow calculated from stream ﬂow loss. They could not determine if that ﬁnding was caused by the difﬁculties in estimating the stream reach surface or due to the travel-time procedure. Subsurface ﬂow rates for depths between 30 and 300 cm calculated from temperature data were highly variable and ranged from 6 to 26 cm h1. Their velocities were obtained in a loosing reach and are consequently much higher than vertical velocities found at the Lahn ﬁeld site. 4.1.2.

Temperature damping

The temperature-damping approach performs well, if material properties are known. Vertical velocity components calculated with this method showed good accordance with the model (error between 7 and 20%). Nevertheless, temperature damping calculated by Taniguchis method (1993) (as described above) showed clearly its strict limitation to very speciﬁc ﬂow velocities and temperature conditions. In this study it was theoretically possible to calculate velocities in the range of 76.6 cm h1. Table 5.4. Correlation between simulated velocities and calculated velocities from time-lag and temperature damping. Depth (cm)

Time-lag (ne) vf ¼ 3–20 cm h1

Time-lag (CM/CW) vf ¼ 3–20 cm h1

1C-damping (Taniguchi, 1993) vf ¼ 2 to 6.6 cm h1

0–20

Y ¼ 2.23 x – 0.01 R2 ¼ 0.99 Y ¼ 2.40 x R2 ¼ 0.996

Y ¼ 0.93 x – 0.01 R2 ¼ 0.99 Y ¼ 0.9998 x R2 ¼ 0.996

Y ¼ 1.22 x – 0.001 R2 ¼ 0.99 Y ¼ 1.007 x – 0.004 R2 ¼ 0.995

20–50

Evaluating vertical velocities

123

However, due to the limitation of the factor b, every velocity beyond this range was calculated to be this maximum value. Therefore, the method failed to give any result for velocities smaller than 2 cm h1, since the daily amplitude became too small to be detected within the sediment. 4.1.3.

1D-model

The performance of the temperature model was good. Mean error in temperature was 1.7%, varying between 0.5 and 4% of the measured temperature range during the test period. In cases where simulated temperatures showed a time-lag in the falling limb of the measured temperature curves (Fig. 5.5, June and August), this pattern was attributed to currents based on density differences. These differences in density caused by water temperatures at the riverbed being below the temperatures within the riverbed lead to higher downward velocities. We observed a similar effect of enhanced velocities due to a density driven ﬂow component in laboratory experiments with coarse bed material. Thus, velocities determined from simulated temperatures underestimate down-welling when surface temperatures are below subsurface temperatures. 4.1.4.

Rating

The travel-time approach, as well as the temperature-damping method, neglects the absolute temperatures, thus ignoring the change of viscosity with temperature. For a change in temperature from 0 to 101C, dynamic viscosity and hydraulic conductivity change about 30%. For data from the investigated ﬁeld site, diurnal changes in viscosity led to errors of calculated velocities of up to 15%. Seasonal differences in mean temperature of about 91C in the surface water lead to differences in calculated ﬂow velocity of up to 27% between summer and winter. While errors and uncertainties accumulate in the travel-time method and the temperature-damping approach, the 1D-model allows for the incorporation of material properties, including their temperature dependence. Furthermore, the relation between the ABW and velocity (Fig. 5.4a) allows the determination of the accuracy of obtained velocity values. Therefore, among those considered, the 1D-model is the method best suited for determining vertical ﬂow velocities within gravel-bed river sediments. 4.2.

Typical periods

The one-dimensional temperature model showed good performance for the study site (Table 5.5). As shown in Fig. 5.6 the determined values of velocity were all in the range of a few cm h1. They showed a pattern of down-welling at location C-1 at the upstream side of the rifﬂe and a predominantly up-welling at the downstream side of the rifﬂe at location C-2. From previous tracer experiments, a strong horizontal current was expected below 50 cm within the riverbed (Saenger 2002). Such ﬂow parallel to the bed in the direction of the stream showed no signiﬁcant inﬂuence on the temperature distribution with depth in the investigated situations. For an

I. Seydell, B.E. Wawra, U.C.E. Zanke

124

Table 5.5. Mean deviance ABW between simulated and measured temperatures (1C) and deviance relative to measured range ABWR (%). June

C-1 20 cm 50 cm 100 cm C-2 20 cm 50 cm 100 cm

August

September

November

ABW

ABWR

ABW

ABWR

ABW

ABWR

ABW

ABWR

0.44 0.19 0.25

4 2 2

0.32 0.09 0.11

3 1 1

0.24 0.14 0.16

2 1 1

0.13 0.06 0.08

1 0.6 0.7

0.45 0.27 0.23

4 2 2

0.25 0.05 0.17

2 0.5 1

0.29 0.23 0.06

3 2 1

0.1 0.07 0.13

0.9 0.6 1

effective porosity of 20%, the associated downward Darcy’s velocity decreased from 2 to 1 cm h1. Vertical velocities between 1 and 3.7 cm h1 were determined for a tracer experiment conducted in the ﬁeld site at the Lahn in 1998 (Saenger, 2000). Values calculated from temperature data for that period showed generally good accordance with those determined from tracer experiments, except at places, where high velocities are attributed to preferential ﬂow paths in combination with extraction of pore water (Seydell et al., in print). The vertical velocity components determined with the 1D-model in this study are in the same order of magnitude as those calculated from salt tracer experiments by Ingendahl (1999) for the Nette and Bro¨hl streams (Germany). Ingendahl observed a factor 2 increase of time-lag attributed to inﬁltration of ﬁne sediment into the initially clean spawning gravel. For the present study the change in subsurface velocity was attributed to changes in subsurface ﬂow pattern rather than inﬁltration of ﬁnes. In June 2000 with maximum diurnal temperature changes of 51C a clear downwelling at location C-1 and a constant up-welling at C-2 was visible. The mean error in temperature (ABWR) of simulated to measured temperature is largest during this period (4%) at 20 cm depth. This is based on a systematic deviation of temperatures during the falling limb of the diurnal variation due to temperature related density effects where surface water temperatures sank below those in the riverbed. In August, which represents a similar situation, the down-welling was less pronounced and already disappeared at 50 cm depth while the up-welling at C-2 was visible only in the upper 20 cm of the riverbed. Here smaller diurnal temperature differences of only 2.51C and higher ﬂow in the stream lead to a reduced exchange along the rifﬂe. The result was a smaller error in temperature and vertical velocity (Table 5.5, Fig. 5.6). For the September period, the velocity pattern within the rifﬂe changed in quantity and quality. Although the mean temperature deviance is small, the error of velocity in relation to the calculated values is large (Fig. 5.6). Surprisingly, down-welling was visible only at location C-2 in 100 cm depth. The intense temperature damping with depth suggests a strong up-welling. The apparent up-welling was induced by a strong

Evaluating vertical velocities

6

6

2

2

-2

-2

-6

-6

6 v (cm h-1)

C-1 at 20 cm

125

C-1 at 50 cm

6

2

2

-2

-2

-6

-6

6

C-1 at 100 cm

6

2

2

-2

-2

-6

-6 June Aug. Sept. Nov.

C-2 at 20 cm

C-2 at 50 cm

C-2 at 100 cm

June Aug. Sept. Nov.

Figure 5.6. Simulated velocities (middle line of box) with maximum error calculated from ABW (upper and lower line of box). Positive values indicate down-welling, negative values up-welling.

lateral current from the enlarged pond. We also observed this up-welling by measuring hydraulic heads. For the period in November, when water temperatures values increased with depth, down-welling was determined at C-1 with values decreasing with depth and again at location C-2 at 100 cm depth with a value of 5 cm h1. As in September, this is explained by a strong transversal ﬂow between 50 and 150 cm depth. Again, this lateral ﬂow was shown by measurements of hydraulic head. Nevertheless, we need to be careful with this interpretation, since water temperature of the pond was unknown and temperatures at different depths where very close to each other in November. The simulated temperature situations illustrate the potential and the limitations of the 1D-model very well. Model performance can be quantiﬁed by the mean deviance between modelled and measured temperatures, but this deviance needs to be transformed into the accuracy of velocity estimated, respectively the error in velocity.

I. Seydell, B.E. Wawra, U.C.E. Zanke

126

If ﬂow pattern are predominantly two dimensional and surface water temperature variation is small in stream ﬂow direction, using a 1D-model is sufﬁcient. 5.

Conclusion

The numerical model is an appropriate tool to evaluate vertical velocity components within the gravel bed from temperature data. For the investigated medium scale exchange between the stream and the hyporheic zone along a rifﬂe structure, the effect of currents induced by density differences due to temperature differences is negligible. The water exchange rate between the stream and the hyporheic pore space can be determined for the zone with designated exchange between the stream and the hyporheic zone. Moreover, comparison of modelled temperature curves with measured ones allows identiﬁcation of unrealistic situations as well as estimation of the achieved accuracy. Temperature measurements showed that the rifﬂe structure inﬂuences the subsurface ﬂow pattern although the gross hydraulic gradient showed major inﬂuence. If temperature varies with depth within the sediment, the numerical modelling of temperature curves is a powerful tool to determine exchange between the stream and the subsurface water body in a wide variety of streams and settings. If temperature data is combined with measurements of hydraulic head, more detailed information on subsurface structures and ﬂow pattern and their variation in time can be extracted. References Behr, M., 2002. Numerische Modellierung der Stro¨mung und des Stofftransportes im hyporheischen Interstitial. Diplomarbeit, University of Technology Darmstadt, Germany. Benson, I., 1999. Using dye tracer to examine a full-scale model of a salmonid redd. Technical Report, Hydroscope Consulting Ltd. Brunke, M., 2000. Wechselwirkungen zwischen FlieXgewa¨sser und Grundwasser: Bedeutung fu¨r aquatische Biodiversita¨t, Stoffhaushalt und Lebensraumstrukturen. Wasserwirt-schaft 90, 32–37. Carling, P.S., Boole, P., 1996. An improved condumetric standpipe technique for measuring interstitial seepage velocity. Hydrobiologica 135, 3–8. Cartwright, K., 1971. Redistribution of geothermal heat by a shallow aquifer. Geol. Soc. Am. Bull. 82 (11), 3197–3200. Chapman, D.W., 1988. Critical review of variables used to deﬁne effects of ﬁnes in redds of large salmonids. Trans. Am. Fisheries Soc. 117, 1–21. Constanz, J., Thomas, C.L., 1996. The use of streambed temperature proﬁles to estimate the depth, duration and rate of percolation beneath arroyos. Water Resour. Res. 32 (12), 3597–3602. Geist, D.R., 2000. The interaction of groundwater and surface water within fall Chinook salmon spawning areas in the Hanford Reach of the Columbia River. Ground-water/surface-water interaction workshop. United States Environmental Protection Agency, Washington, pp. 95–98. Gibert, J., Dole-Olivier, M.J., Marmonier, P., Vervier, P., 1990. Surface waterground water ecotones. In: Naiman, R.J. and Decamps, H. (Eds), The ecology and management of aquatic terrestital ecotones. Man and the Biosphere Series, Vol. 4. UNESCO, Paris and Parthenon Publishing Group, Carnforth, UK, pp. 199–226. Gwo, J.P., D’Azevedo, E.F., Frenzel, et al., 1999. HydroBioGeoChem123D: A Coupled Model of Hydrologic Transport and Mixed Biogeochemical Kinetic/Equilibrium Reactions in Saturated-Unsaturated

Evaluating vertical velocities

127

Media in One, Two, and Three Dimensions. Computer code and documentation available at http:// hbgc.esd.ornl.gov Hakenkamp, C.C., Palmer, M.A., 2000. The ecology of hyporheic meiofauna. In: Jones, J. and Mulholland, P. (Eds), Streams and ground waters. Academic Press, San Diego, pp. 307–336. Harvey, J.W., Wagner, B.J., Bencala, K.E., 1996. Evaluating the reliability of the stream tracer approach to characterize stream-subsurface water exchange. Water Resour. Res. 32, 2441–2451. Ingendahl, D., 1999. Der Reproduktionserfolg von Meerforelle (Salmo trutta L.) und Lachs (Salmo salar L.) in Korrelation zu den Milieubedingungen des hyporheischen Interstitials. Dissertation. Institut fu¨r Zoologie, Universita¨t zu Ko¨ln, Germany. Ingendahl, D., 2001. Dissolved oxygen concentration and emergence of sea trout fry from natural redds in tributaries of the River Rhine. J. Fish Biol. 58, 325–341. Kirchmaier, N., 2003. Numerische Modellierung der Temperaturen im hyporheischen Interstitial. Institut fu¨r Wasserbau und Wasserwirtschaft, University of Technology Darmstadt, Germany. Report, unpublished. Lenk, M., 2000. Hydraulische Austauschvorga¨nge zwischen FlieXgewa¨sser und Interstitial – Felduntersuchungen in einer Pool-Rifﬂe-Sequenz an der oberen Lahn. Dissertation, Wasserbauliche Mitteilungen des Instituts fu¨r Wasserbau und Wasserwirtschaft, Heft 114, TU Darmstadt. Malcom, I.A., Soulsby, A.F., Youngson, D.M., et al., 2004. Hydrological inﬂuences on hyporheic water quality: implications for salmon egg survival. Hydrol. Process. 18, 1543–1560. Nash, J.E., Sutcliffe, J.V., 1970. River ﬂow forecasting through conceputal models Part1 – a discussion of principles. J. Hydrol. 10, 282–290. Petticrew, E.L., Kalff, J., 1991. Calibration of a gypsum source for freshwater ﬂow measurements. Can. J. Fisheries Aquatic Sci. 48, 1244–1249. Pusch, M., 1993. Heterotropher Stoffumsatz und faunistische Besiedlung des hyporheischennterstitials eines Mittelgebirgsbaches (Steina, Schwarzwald). – Dissertation, Universita¨t Freiburg. Rauch, W., 1991. Ausbreitung von Temperaturanomalien im Grundwasser. Dissertation, Baufakulta¨t, Universita¨t Insbruck, Austria. Saenger, N., 2002. Estimation of ﬂow velocity within the hyporheic zone. Verhandlungen der Internatonalen Vereinigung fu¨r theoretische und angewandte Limnologie 28 (4), 1790–1795. Saenger, N., Zanke, U.C.E. (in print). A depth-oriented view of hydraulic exchange patterns between surface water and the hyporheic zone – Analysis of ﬁeld experiments at the River Lahn, Germany. Archiv fu¨r Hydrobiologie. Saenger, N., 2000. Identiﬁkation von Austauschprozessen zwischen FlieXgewa¨sser und hyporheischer Zone. Wasserbauliche Mitteilungen, 114. Institut fu¨r Wasserbau und Wasserwirtschaft, University of Technology Darmstadt, Germany. Scheffer, F., Schachtschabel, P., 2002. Lehrbuch der Bodenkunde. Vol. 15. Spektrum Akademischer Verlag, Auﬂ. Heidelberg. Seydell, I., Wawra, B., Zanke, U.C.E., (in print). Patterns of permeability and clogging processes in the hyporheic zone of a gravel bed river (River Lahn, Germany). Archiv fu¨r Hydrobiologie. Silliman, S.E., Booth, D.F., 1993. Analysis of time series measurements of sediment temperature for identiﬁcation of gaining versus losing portions of Juday Creek, Indiana. J. Hydrol. 146, 146–148. Stallmann, R.W., 1965. Steady one-dimensional ﬂuid ﬂow in semi-inﬁnite porous medium with sinusoidal surface temperature. J. Geophys. Res. 70, 2821–2827. Storey, R.G., Howard, W.F., Williams, D.D., 2003. Factors controlling rifﬂe-scale exchange ﬂows and their seasonal change in a gaining stream: A three-dimensional groundwater ﬂow model. Water Resour. Res. 39 (2), Art. No. 1034, FEB 18 2003. Taniguchi, M., 1993. Evaluation of vertical groundwater ﬂuxes and thermal properties of aquifers based on transient temperature–depth proﬁles. Water Resour. Res. 29 (7), 2021–2026. Thompson, L.T., Glenn, E.P., 1994. Plaster standards to measure water motion. Limnol. Oceanogr. 39 (1), 1768–1778. Vervier, P., Gibert, J., Marmonier, P., Dole-Olivier, M.-J., 1992. A perspective on the permeability of the surface freshwater–groundwater ecotone. J. N Am. Benthol. Soc. 11 (1), 92–102.

128

I. Seydell, B.E. Wawra, U.C.E. Zanke

Discussion by Ian Reid Seydell et al. provide an interesting analysis of inferred ﬂow within the hyporheus of a gravel bar by using thermal properties of the interstitial water as a ‘tracer’ to calibrate their model. However, the analysis appears to presume that the properties of the porous media constituting the river bed are isotropic. The stratigraphy of both ancient (e.g., Nemec and Steel, 1984) and modern (e.g., Hassan, 2005; Laronne and Shlomi, 2007) gravel-bed rivers shows us that superjacent beds with signiﬁcantly different grain-size and, hence, pore-size distributions vary considerably in thickness and 3D geometry at decimetre scale. While other studies of matrices show large variations in the degree to which pores are ﬁlled and give a rationale for understanding the vertical juxtaposition of open-framework and matrix-ﬁlled interstices even within the same bed (e.g., Frostick et al., 1984; Reid and Frostick, 1985). These studies suggest that gravel-bed rivers are more likely than not to be characterized by anisotropy of hydraulic conductivity, both vertical and horizontal, and this must have signiﬁcant impact on interstitial ﬂow velocity. Questions arise, therefore, as to whether properties and dimensions of gravel-bed facies need to be taken into account when modelling hyporheic ﬂow nets? or whether abrupt changes in hydraulic conductivity are inconsequential in affecting ﬂows as small as those reported by Seydell et al. (up to 14 mm s1), except, presumably, where pores are tightly packed with ingressed clay grains or a clay drape is interdigitated in the microstratigraphy? References Frostick, L.E., Lucas, P.M., Reid, I., 1984. The inﬁltration of ﬁne matrices into coarse-grained alluvial sediments and its implications for stratigraphical interpretation. J. Geol. Soc. London 141, 955–965. Hassan, M.A., 2005. Characteristics of gravel bars in ephemeral streams. J. Sediment. Res. 75, 203–221. Laronne, J.B., Shlomi, Y., 2007. Depositional character and preservation potential of coarse grained sediments deposited by ﬂood events in hyper-arid braided channels in the Rift Valley, Arava, Israel. Sediment. Geol. 195(1–2), 21–37. Nemec, W., Steel, R.J., 1984. Alluvial and coastal conglomerates: their signiﬁcant features and some comments on gravely mass-ﬂow deposits. In: Koster, E.H. and Steel, R.J. (Eds), Sedimentology of gravels and conglomerates. Memoir Canadian Society Petroleum Geologists 10, 1–31. Reid, I., Frostick, L.E., 1985. Role of settling, entrainment and dispersive equivalence and of interstice trapping in placer formation. J. Geol. Soc. London 142, 739–746.

Discussion by John M. Bufﬁngton1 & Daniele Tonina2 Hyporheic exchange is principally driven by spatial variations of near-bed pressure, which can be sensitive to seasonal changes in discharge, ﬂow depth, and watersurface proﬁle (Tonina and Bufﬁngton, 2003, 2007). Simulations of hyporheic exchange across two-dimensional pool-rifﬂe topography show that the strength and spatial extent of the hyporheic exchange vary with discharge (Fig. 5.7). High 1

US Forest Service, Rocky Mountain Research Station, 322 E Front St., Boise, Idaho 83702, USA. US Forest Service, Rocky Mountain Research Station, 322 E Front St., Boise, Idaho 83702, USA and Department of Earth & Planetary Science, University of California, Berkeley, California 94720, USA. 2

128

I. Seydell, B.E. Wawra, U.C.E. Zanke

Discussion by Ian Reid Seydell et al. provide an interesting analysis of inferred ﬂow within the hyporheus of a gravel bar by using thermal properties of the interstitial water as a ‘tracer’ to calibrate their model. However, the analysis appears to presume that the properties of the porous media constituting the river bed are isotropic. The stratigraphy of both ancient (e.g., Nemec and Steel, 1984) and modern (e.g., Hassan, 2005; Laronne and Shlomi, 2007) gravel-bed rivers shows us that superjacent beds with signiﬁcantly different grain-size and, hence, pore-size distributions vary considerably in thickness and 3D geometry at decimetre scale. While other studies of matrices show large variations in the degree to which pores are ﬁlled and give a rationale for understanding the vertical juxtaposition of open-framework and matrix-ﬁlled interstices even within the same bed (e.g., Frostick et al., 1984; Reid and Frostick, 1985). These studies suggest that gravel-bed rivers are more likely than not to be characterized by anisotropy of hydraulic conductivity, both vertical and horizontal, and this must have signiﬁcant impact on interstitial ﬂow velocity. Questions arise, therefore, as to whether properties and dimensions of gravel-bed facies need to be taken into account when modelling hyporheic ﬂow nets? or whether abrupt changes in hydraulic conductivity are inconsequential in affecting ﬂows as small as those reported by Seydell et al. (up to 14 mm s1), except, presumably, where pores are tightly packed with ingressed clay grains or a clay drape is interdigitated in the microstratigraphy? References Frostick, L.E., Lucas, P.M., Reid, I., 1984. The inﬁltration of ﬁne matrices into coarse-grained alluvial sediments and its implications for stratigraphical interpretation. J. Geol. Soc. London 141, 955–965. Hassan, M.A., 2005. Characteristics of gravel bars in ephemeral streams. J. Sediment. Res. 75, 203–221. Laronne, J.B., Shlomi, Y., 2007. Depositional character and preservation potential of coarse grained sediments deposited by ﬂood events in hyper-arid braided channels in the Rift Valley, Arava, Israel. Sediment. Geol. 195(1–2), 21–37. Nemec, W., Steel, R.J., 1984. Alluvial and coastal conglomerates: their signiﬁcant features and some comments on gravely mass-ﬂow deposits. In: Koster, E.H. and Steel, R.J. (Eds), Sedimentology of gravels and conglomerates. Memoir Canadian Society Petroleum Geologists 10, 1–31. Reid, I., Frostick, L.E., 1985. Role of settling, entrainment and dispersive equivalence and of interstice trapping in placer formation. J. Geol. Soc. London 142, 739–746.

Discussion by John M. Bufﬁngton1 & Daniele Tonina2 Hyporheic exchange is principally driven by spatial variations of near-bed pressure, which can be sensitive to seasonal changes in discharge, ﬂow depth, and watersurface proﬁle (Tonina and Bufﬁngton, 2003, 2007). Simulations of hyporheic exchange across two-dimensional pool-rifﬂe topography show that the strength and spatial extent of the hyporheic exchange vary with discharge (Fig. 5.7). High 1

US Forest Service, Rocky Mountain Research Station, 322 E Front St., Boise, Idaho 83702, USA. US Forest Service, Rocky Mountain Research Station, 322 E Front St., Boise, Idaho 83702, USA and Department of Earth & Planetary Science, University of California, Berkeley, California 94720, USA. 2

Evaluating vertical velocities

129

Figure 5.7. Simulated hyporheic pathlines for a synthetic two-dimensional pool-rifﬂe topography with (a) low discharge (8% bankfull ﬂow) and (b) high discharge (100% bankfull ﬂow). Channel characteristics scaled from natural gravel-bed rivers in central Idaho: slope is 0.5% and the ratio of bedform amplitude to wavelength, D/l, is 0.03. Surface and subsurface ﬂow simulated using MD_SWMS 1 (McDonald et al., 2005) and FLUENT 6.0 (FLUENT Inc.), respectively. Predicted water surface proﬁles are plotted separately above each panel at exaggerated vertical scales to emphasize their differences, but do not indicate ﬂow depths over bed topography. Subsurface ﬂow simulations use a uniform hydraulic conductivity of 0.1 cms1 (sandy gravel), with lighter-coloured hyporheic pathlines indicating faster ﬂow.

130

I. Seydell, B.E. Wawra, U.C.E. Zanke

Figure 5.8. Simulated hyporheic pathlines for a synthetic three-dimensional pool-rifﬂe channel with alternate bar morphology and low discharge (26% bankfull ﬂow). Channel characteristics scaled from natural gravel-bed rivers in central Idaho: slope is 0.41% and the ratio of bedform amplitude to wavelength is 0.022 (Tonina and Bufﬁngton, 2007, experiment 1). Surface and subsurface ﬂow simulated using FLUENT 6.0 (FLUENT Inc.), with hyporheic pathlines originating from the surface and coloured by individual trajectory. The simulation uses a groundwater slope of 0.41%, a uniform hydraulic conductivity of 5 cm s1 (Tonina and Bufﬁngton, 2007), and an alluvial depth equal to one bedform wavelength (1l). Deeper groundwater ﬂow paths (those that do not intersect the bed surface) are not shown.

Evaluating vertical velocities

131

discharges decrease the magnitude of hyporheic exchange in pool-rifﬂe channels by smoothing the water-surface proﬁle and decreasing the spatial variation of near-bed pressure (Fig. 5.7) (Tonina and Bufﬁngton, 2007). Furthermore, the direction of hyporheic ﬂow (up-welling vs. down-welling) can change with discharge. For example, just downstream of the rifﬂe crest, hyporheic ﬂow is predicted to up-well at low discharge, but down-well at high discharge (Fig. 5.7). These changing patterns and magnitudes of hyporheic exchange together with the authors’ limited number of ﬁxed sample sites may partially explain the seasonal variations in hyporheic ﬂow observed at their study site (their Fig. 5.6). Furthermore, hyporheic ﬂow paths are even more complex and strongly three dimensional in pool-rifﬂe channels with alternate bar morphology (Fig. 5.8). Interactions between ﬂow and bedform topography induce lateral hyporheic ﬂow in these channels that can change with discharge and may contribute to the lateral ﬂow observed by the authors.

References McDonald, R.R., Nelson, J.M., Bennett, J.P. 2005. Multi-dimensional surface-water modeling system user’s guide. U.S. Geological Survey Techniques and Methods, 6-B2, 136pp. (http://wwwbrr.cr.usgs. gov/projects/SW_Math_mod/OpModels/MD_SWMS/index.htm) Tonina, D., Bufﬁngton, J.M., 2003. Effects of discharge on hyporheic ﬂow in a pool-rifﬂe channel: Implications for aquatic habitat. EOS Trans. Am. Geophys. Union 84 (46), Fall Meeting Supplement, Abstract H52A-1154. Tonina, D., J.M. Bufﬁngton., 2007. Hyporheic exchange in gravel-bed rivers with pool-rifﬂe morphology: Laboratory experiments and three-dimensional modeling. Water Resources Research, 43, W01421.

Reply by the authors We thank the discussants for highlighting important points. The described changes in hyporheic ﬂow are greatly attributed to in channel ﬂow and do not reﬂect necessarily seasons but different ﬂow conditions. In the Lahn ﬁeld site one interesting ﬁnding was the change from seemingly up-welling in the upper 50 cm of sediment, to downwelling at 100 cm depth during September 2000. Here the additional inﬂuence of ground water level and, in the investigated case the induced strong lateral velocity component is visible. This points out the limitation of the one-dimensional approach main aim of which is to provide an simple method for research interested in only limited information on vertical exchange between the stream and the pore water. Beyond this it is most desirable to have a 3D model as described by the Discussants with additional temperature data for a natural river bed, to determine spatial heterogeneity with greater reliability. This is currently done by the Department of Hydrogeology, UFZ Centre for Environmental Research, Leipzig-Halle, Germany.

Evaluating vertical velocities

131

discharges decrease the magnitude of hyporheic exchange in pool-rifﬂe channels by smoothing the water-surface proﬁle and decreasing the spatial variation of near-bed pressure (Fig. 5.7) (Tonina and Bufﬁngton, 2007). Furthermore, the direction of hyporheic ﬂow (up-welling vs. down-welling) can change with discharge. For example, just downstream of the rifﬂe crest, hyporheic ﬂow is predicted to up-well at low discharge, but down-well at high discharge (Fig. 5.7). These changing patterns and magnitudes of hyporheic exchange together with the authors’ limited number of ﬁxed sample sites may partially explain the seasonal variations in hyporheic ﬂow observed at their study site (their Fig. 5.6). Furthermore, hyporheic ﬂow paths are even more complex and strongly three dimensional in pool-rifﬂe channels with alternate bar morphology (Fig. 5.8). Interactions between ﬂow and bedform topography induce lateral hyporheic ﬂow in these channels that can change with discharge and may contribute to the lateral ﬂow observed by the authors.

References McDonald, R.R., Nelson, J.M., Bennett, J.P. 2005. Multi-dimensional surface-water modeling system user’s guide. U.S. Geological Survey Techniques and Methods, 6-B2, 136pp. (http://wwwbrr.cr.usgs. gov/projects/SW_Math_mod/OpModels/MD_SWMS/index.htm) Tonina, D., Bufﬁngton, J.M., 2003. Effects of discharge on hyporheic ﬂow in a pool-rifﬂe channel: Implications for aquatic habitat. EOS Trans. Am. Geophys. Union 84 (46), Fall Meeting Supplement, Abstract H52A-1154. Tonina, D., J.M. Bufﬁngton., 2007. Hyporheic exchange in gravel-bed rivers with pool-rifﬂe morphology: Laboratory experiments and three-dimensional modeling. Water Resources Research, 43, W01421.

Reply by the authors We thank the discussants for highlighting important points. The described changes in hyporheic ﬂow are greatly attributed to in channel ﬂow and do not reﬂect necessarily seasons but different ﬂow conditions. In the Lahn ﬁeld site one interesting ﬁnding was the change from seemingly up-welling in the upper 50 cm of sediment, to downwelling at 100 cm depth during September 2000. Here the additional inﬂuence of ground water level and, in the investigated case the induced strong lateral velocity component is visible. This points out the limitation of the one-dimensional approach main aim of which is to provide an simple method for research interested in only limited information on vertical exchange between the stream and the pore water. Beyond this it is most desirable to have a 3D model as described by the Discussants with additional temperature data for a natural river bed, to determine spatial heterogeneity with greater reliability. This is currently done by the Department of Hydrogeology, UFZ Centre for Environmental Research, Leipzig-Halle, Germany.

Gravel-Bed Rivers VI: From Process Understanding to River Restoration H. Habersack, H. Pie´gay, M. Rinaldi, Editors r 2008 Elsevier B.V. All rights reserved.

133

6 Bifurcations in gravel-bed streams Marco Tubino and Walter Bertoldi

Abstract In the present paper we provide an overview on some recent experimental and theoretical works, which have been specifically designed to analyze the behaviour of bifurcations in gravel-bed streams. We ﬁrst investigate the occurrence of a bifurcation starting from an initially straight channel: the interaction between bed and banks evolution determines ﬂow bifurcation when channel width oscillations reach a maximum amplitude; the process is strongly dependent on the migration speed of bars that form in the channel. Experimental evidence on the equilibrium conﬁgurations and stability of a simple Y-shaped bifurcation, both in the case of ﬁxed and erodible-bank channels, shows that bifurcations are likely to display unbalanced conﬁgurations, characterized by uneven partition of ﬂow discharge and different values of free surface width of downstream branches. Moreover bed levels of downstream branches differ, in accordance with ﬁeld observations on gravel-bed rivers. Theoretical predictors based on 1D schemes supplemented by suitable 2D information at the node satisfactorily replicate most of the observed features. When the channels joining at the node are free to evolve, the stability of bifurcations is controlled by the occurrence of migrating bars and by the adaptation processes of channel width and planform, as well by the morphodynamic inﬂuence of the bifurcation. The resulting evolution may also depend on the initial mechanism triggering ﬂow bifurcation.

1.

Introduction

Braided rivers are intrinsically dynamical systems, characterized by rapid and frequent changes of bed topography and planform shape. They are typically unsteady, never reaching a static equilibrium conﬁguration, not even in the case of a controlled laboratory model with constant water and sediment supply. They display a hierarchy of spatial and temporal scales, which range from the small scale typical of sediment transport adaptation processes and bedforms development to the intermediate-large E-mail address: [email protected] (M. Tubino) ISSN: 0928-2025

DOI: 10.1016/S0928-2025(07)11123-8

134

M. Tubino, W. Bertoldi

scales reﬂecting the dynamics of single branches and nodes and of the whole braided belt. Such scales are also determined by historical legacies related to sedimentological processes and to the variations of hydrological regime of ﬂow and sediment supply (Ashmore, 1991; Hoey, 1992; Ashmore, 2001). Furthermore, in active gravel-bed networks bank cohesion and vegetation are likely to play a minor, albeit complementary, role (e.g., Gran and Paola, 2001; Gurnell et al., 2001). These evolutionary processes are strongly interconnected and the above scales range on a continuum where a clear distinction among different scales can hardly be established. Such a complexity is typically reﬂected in the records of sediment transport collected at the downstream-end of braided network laboratory models (Ashmore, 1988; Hoey and Sutherland, 1991; Warburton and Davies, 1994). The analysis of these evolutionary scales is a fundamental step to improve our understanding of the behaviour of river systems and of their response to natural and anthropogenic controls. This investigation gains even more relevance when considering that recent approaches to ﬂuvial management look at a braided pattern as a suitable context to enhance different river functionalities. In fact, its intrinsic properties determine a dynamic environment with a high ecosystem diversity; furthermore, they lead to a system which displays a greater adaptability to changing ﬂow conditions and can also mitigate the effect of extreme events (Gilvear, 1999; Klaassen et al., 2002; Van der Nat et al., 2002). In spite of the inherent complexity of braided systems, it proves to be instructive to single out the various constitutive processes and to examine their dynamics independently. The above viewpoint is adopted herein, where speciﬁc attention is devoted to the analysis of the bifurcation process, whose evolution constitutes a major ingredient responsible for the complexity of gravel-bed braided rivers (Fig. 6.1 shows an example of a bifurcation in a gravel-bed braided river). In fact, the partition of water and sediment discharge in the active channels of a braided network is mainly controlled by the occurrence of bifurcations and by their subsequent development (Leopold and Wolman, 1957; Ashmore, 1991, Bristow and Best, 1993).

Figure 6.1. Bifurcation in a gravel-bed braided river (Tagliamento River, Italy).

Bifurcations in gravel-bed streams

135

Field observations of Leopold and Wolman (1957) have ﬁrst highlighted the role of bifurcations as the primary cause of braiding. Further investigations have widely documented the occurrence of chute and lobe units in gravel-bed braided rivers and the close relationship between bifurcations and conﬂuence–difﬂuence units, along with the control exerted on local bed topography and discharge partition in downstream branches by sediment transport processes (e.g., Southard et al., 1984; Ferguson et al., 1992). On a much larger spatial scale and in a different context, Richardson and Thorne (2001) have described the ﬂow ﬁeld in newly formed bifurcations on the sandy Jamuna River, Bangladesh. The above investigations suggest that the onset of a bifurcation can be mainly, though not invariably, related to the occurrence of chute cutoffs due to local ﬂow acceleration. The frequent occurrence of chute cutoffs in braided streams is also reﬂected by the weakly meandering character of single distributaries within the network (e.g., Krigstrom, 1962; Klaassen and Masselink, 1992). This picture has been also conﬁrmed by several investigations carried out in laboratory models of braided networks (e.g., Ashmore, 1991), though other processes have been also identiﬁed through which bifurcations may set the onset of braiding, namely central bar initiation and emergence (Ashworth, 1996), transverse bar conversion (chute and lobe), multiple bar braiding and partial avulsion (Slingerland and Smith, 2004). Further insight into the distinctive features displayed by gravel-bed bifurcations in natural contexts has been recently gained through the observations noted by Zolezzi et al. (2006) in two series of ﬁeld campaigns on the Sunwapta River (Alberta, Canada) and on the Ridanna Creek (South Tyrol, Italy). The results of the detailed survey of the morphometric and hydraulic characteristics of seven selected bifurcations within the two study reaches suggest that the main recurring feature is a strong asymmetry of the morphological conﬁguration at the node, which invariably results in an uneven partition of ﬂow and sediment transport in downstream branches. The branch carrying the larger discharge is found to be invariably wider and deeper. Such unbalanced distribution of ﬂow discharge is mainly driven by topographical effects: local aggradation establishes just upstream of the bifurcation point, due to ﬂow divergence, followed by a sudden bed degradation at the inlet of the larger branch. As a result, a transverse bottom inclination characterizes the bifurcation region, whose effect may extend over a length of few channel widths upstream of the bifurcation point; furthermore, an inlet step establishes, whereby the bed level at the entrance of the smaller downstream branch sits at a higher elevation than the other branch. Field observations of Zolezzi et al. (2006) also suggest that the above asymmetry often implies that only the larger branch is morphologically active, at least at low-intermediate stages; furthermore, discharge anomaly is found to increase for decreasing values of the incoming discharge. The question then arises on whether the above asymmetry is inherently associated with the dynamic behaviour of the bifurcation or it mainly results from the combined role of non-local inﬂuences. The latter can be exerted, among others, by the migration of bars in the upstream channel, by the adjustment process to the ﬂowing discharge of the width and curvature of channels joining at the node and by the occurrence of backwater effects. In principle, we may expect that a bifurcation remains stable when the transport capacity of downstream branches balances their

136

M. Tubino, W. Bertoldi

load at the mouth; otherwise, local sedimentation and erosion occur and the system undergoes morphological changes. However, both the incoming load and the sediment carrying capacity of downstream branches depend on local ﬂow conditions and bed topography at the node as well as on non-local effects due to external forcings. On one hand one could argue that the sudden deviation of the main ﬂow orientation induced by a bifurcation may act as a planimetric discontinuity, like that associated with an abrupt change of channel curvature, leading to a morphodynamic inﬂuence on bed topography whose effect may be felt over a distance of several channel widths. In this respect the avenue opened by the work of Zolezzi and Seminara (2001) may provide a sound explanation of the observed behaviour of bifurcations, as it states that the inﬂuence of such a discontinuity is mainly felt upstream or downstream depending on the aspect ratio b falling above or below a resonant value br, which depends on sediment mobility and grain roughness. We note that the concept of resonance in river morphodynamics has been originally introduced by Blondeaux and Seminara (1985) (see also Seminara and Tubino, 1992), who have shown that the linear solution for ﬂow and bed topography in meanders with a periodic distribution of channel curvature exhibits a resonant behaviour as b approaches br. For values of relevant parameters typical of single distributaries in gravel-bed braided rivers, the above resonant value is relatively small and falls within the range (5C15)1. Hence, bifurcations displaying a super-resonant behaviour can be frequently encountered in gravel-bed braided rivers, which would imply a dominant upstream inﬂuence. On the other hand, ﬁeld and laboratory observations suggest that the dynamics of channels and nodes are strongly related. Channel adjustment is required by the processes of node shifting, creation or annihilation; in turn, channel migration affects the bifurcation. Furthermore, bifurcations often occur after a well-deﬁned sequence of in-channel events which reﬂect the strong interaction between the planimetric and the altimetric conﬁgurations (Ashmore, 1991). The latter processes have been given a sound explanation through a large number of theoretical and experimental works (for a detailed review on the subject see Bolla Pittaluga et al., 2001), though the present knowledge is almost exclusively restricted to the case of ﬁxed-bank channels. We also note that the available experimental data on channel bifurcations mainly provide a static description of the system, like that commonly used at a larger scale to deﬁne the overall properties of a network (link length, braiding indexes) (Ashmore, 1991). On the contrary, the assessment of the relative role of the various ingredients, which affect the stability and the evolution of the nodes would require the availability of detailed records of evolutionary processes and the consequent identiﬁcation of the corresponding timescales. The lack of this information is the main reason why predicting the behaviour of a bifurcation is still a challenge for existing mathematical models, even for simple conﬁgurations (Klaassen et al., 2002; Jagers, 2003). In this respect we observe that the short time evolution of braided networks often seems to be mainly determined by the morphodynamic processes occurring in few, simultaneously active channels. Due to the generally observed asymmetry of the 1

b is the ratio between the half-width of the channel and the ﬂow depth.

Bifurcations in gravel-bed streams

137

bifurcations, these branches carry most of the water discharge, while the rest of the network is poorly affected by sediment transport (Mosley, 1983; Stojic et al., 1998). Such considerations justify, on one hand, the increasing interest for the development of simpliﬁed models, which are suitably designed to reproduce the network evolution at a ‘channel scale’ (Paola, 2001; Jagers, 2003), and, on the other hand, claim for a sounding characterization of single loops within the network. In the present paper we provide a summary of some recent experimental and theoretical works, which have been specifically designed to analyze the behaviour of a single bifurcation. We ﬁrst investigate the process that leads to the establishment of a bifurcation within individual branches. Then we analyze the equilibrium conﬁgurations of a simple Y-shaped bifurcation, along with their stability, in the case of ﬁxedbank channels. Finally, we leave the planform of the channels to be free to evolve and investigate the control exerted on the stability of the bifurcation by the occurrence of migrating bars and by the adaptation process of the channel width and planform.

2.

The onset of braiding through chute cutoff

Since the original work of Friedkin (1945) an analogous experimental procedure has been adopted by various authors to investigate the onset of braiding. An initially straight narrow channel of trapezoidal shape is cut into a cohesionless ﬂat sloping surface and fed with a constant discharge. After some time, such a conﬁguration invariably breaks up into a multichannel pattern, following a sequence of events resulting from the interaction between bed and channel processes. The same procedure has been recently used by Bertoldi and Tubino (2005) (hereinafter referred to as BT) to provide a detailed quantitative analysis of the processes which lead to the establishment of chute cutoffs in individual channels evolving to a bifurcated state. They have performed two series of experiments using well-sorted quartz sand distributions, with values of the mean diameter Ds of 0.5 and 1.3 mm, respectively, and two further series using bimodal mixtures (0:5C1:3 mm; 0:5C1:9 mm) with equal percentages of the two fractions. In terms of the initial values of relevant dimensionless parameters, namely the aspect ratio b and the Shields stress W, the experimental conditions of BT cover the range b0 ¼ ð3:1C6Þ and W0 ¼ ð0:07C0:16Þ, W¼

tn ; ðrs rÞgDs

b¼

b , 2D

(6.1a,b)

where b is the free surface width, D the reach averaged value of water depth, rs and r the sediment and water densities, respectively, g the acceleration due to gravity and t* the average bed shear stress (the subscript 0 denotes the initial values of parameters). BT have carefully documented the planimetric and altimetric development of the channel through series of pictures taken from a digital camera, which could be moved along the longitudinal direction, and frequent surveys of bed topography, performed through a laser scanning device. In such a way they have measured the bed and channel conﬁguration at different stages, from the initial state characterized

138

M. Tubino, W. Bertoldi

by the development of alternate bars in a nearly straight channel, the subsequent development of a weakly meandering channel displaying large width oscillations, until the ﬁnal occurrence of ﬂow bifurcations through chute cutoffs. The above observations concern a highly simpliﬁed bifurcating system, where the unsteady character of water and sediment inputs and the reworking effects resulting from the interaction between different branches are not considered; however, they provide a physical basis to set suitable rules which can be used in predictive models of channel changes in braided rivers (Jagers, 2003). Experimental results of BT suggest that the bed pattern of individual channels is highly reworked during the above evolutionary sequence, though some distinctive characteristics of the initial morphodynamical processes persist even in the ﬁnal stage. In particular, the planimetric non-uniformities of the channel (curvature and width variations), whose intensity gradually increases, strongly affect bed topography and promote the transition from migrating free responses, which mostly exhibit an alternate bar structure, to a quasi-steady conﬁguration where central bar structures (or more complex transverse structures) eventually dominate. This clearly results from the Fourier spectra of bed topography, as shown in Fig. 6.2a where the amplitude of the leading components of bed topography measured at the initial stage (open symbols) is compared with that measured immediately before ﬂow bifurcation (closed symbols) for the whole set of experiments performed by BT. On the other hand, the longitudinal spacing of bifurcation points is found to be essentially controlled by the length of bars that form in the initial stage of the process and lead to a sequence of erosional bumps along both banks. In fact, both the length of bars and that of bank oscillations remain almost ﬁxed during further evolution of the channel. This fact has been also conﬁrmed by further laboratory observations performed in channels developing within a laboratory model of a braided network (Bertoldi et al., 2005). Due to the strong interrelation between the altimetric and the planimetric pattern, the length of the bars which form in newly developing channels is set by the initial formative process, and it is then unable to conform to further channel widening. This may be seen as an indication of the fact that the length scale of morphological processes loses its dependence on the actual channel width, which, on the contrary, would be implied by the results of bed stability theories for ﬁxed-bank channels (e.g., Colombini et al., 1987). Further in general, the experimental results of BT suggest that linear theories, which have led to the establishment of theoretical predictors of river planform based on bar formation criteria (e.g., Fredsoe, 1978; Kuroki and Kishi, 1985) can hardly be applied to predict the transverse mode selection which sets the onset of braiding in gravel-bed bifurcating channels, since alternate bar can undergo a ﬁnite amplitude development which is strongly conditioned by non-linear effects resulting from the forcing effect of planform shape. Experimental observations of BT also indicate that the control exerted by the latter forcing on the subsequent channel dynamics is crucially dependent on the ratio between the migration speed of bars and the bank erosion rate, such that a quite different scenario can be observed depending on local ﬂow conditions. At relatively high values of sediment mobility, bar migration is sufﬁciently fast to prevent the occurrence of intense localized bank erosion; this, in turn, inhibits the ampliﬁcation

Bifurcations in gravel-bed streams

139

amplitude of modes 2 + 3

a)

b)

0.5

0.35

max 0.3

initial stage incipient bifurcation

0.4

0.25

0.3

0.2 0.2

0.15

0.1

0.1

0 0

0.1

0.2

0.3

0.4

0.5

amplitude of mode 1

c) erosion rate [m/s]

0.05 0.06

0.07

d)

1

umax

downstream migration lateral migration

0.08 Shields stress

0.09

0.1

0.04 c

1

0.8

0.1

0.03

0.6 0.02 0.4

0.01

0.2 0.001 0.06

0.07

0.08

0.09

0.10

0.11

Shields stress

e) angle of bifurcation [˚]

uniform sediments graded sediments fast runs

0.07

0.08

0.09

0 0.11

0.10

Shields stress

f)

70

0.3 0.25

uniform sediments 60

0 0.06

0.01

umax migration speed

graded sediments

0.2

50

0.15

40

0.1

30

0.05 0

20 10

15

20 aspect ratio

25

30

5

10

15

20 25 30 aspect ratio

35

40

Figure 6.2. Experimental results of Bertoldi and Tubino (2005): (a) the amplitude of alternate bars (mode 1) and of central and multiple bars (mode 2 and 3) at the initial stage and at the bifurcation; (b) peak values of the dimensionless amplitude of bank oscillations dmax , scaled with half channel width, as a function of Shields stress (crosses denote those runs where bars were not observed to cease their migration); (c) downstream and lateral migration of the channel as a function of Shields stress for the experimental runs B; (d) the migration speed of bars, c, and the local excess of longitudinal velocity at the bank, umax , scaled by the mean ﬂow velocity, as predicted by linear stability analysis of free bars (Colombini et al., 1987) for the experimental runs B; (e) bifurcation angles; (f) the dimensionless amplitude of bank oscillations d, scaled with half channel width, is plotted as a function of the aspect ratio.

of channel curvature and of width variations, whose intensity remains small and is then unable to reduce significantly bar migration. At lower values of Shields stress channel planform develops on a timescale comparable with that of bar migration, which implies that the amplitude of bank oscillations may reach larger values, as shown in Fig. 6.2b. Under these conditions, both channel curvature and width variations significantly slow down bar migration speed, such that bars can no longer move relative to the channel. In this respect the observed phenomena conform to the results of previous laboratory and theoretical investigations on ﬁxed-bank channels with variable curvature (Kinoshita and Miwa, 1974; Tubino and Seminara, 1990) and variable width (Repetto and Tubino, 1999; Repetto et al., 2002).

140

M. Tubino, W. Bertoldi

We note that both the migration speed and the bank erosion rate are essentially related to bar dynamics, since in laterally unconstrained channels the rate of local erosion is mainly controlled by the topographical expression of migrating bars. Hence, the observed dependence on Shields parameter of the evolutionary process of bifurcating channels, which is summarized in Fig. 6.2c for one series of BT’s experiments, can be given as an explanation on the basis of theoretical results for alternate bars (e.g., Seminara and Tubino, 1989). In fact, such results suggest that bar migration becomes faster as Shields stress increases, while the maximum value of the excess velocity at the banks due to the presence of bars, which provides a suitable measure of the rate of bank retreat, remains almost constant (see Fig. 6.2d). It is worth noticing that, though the resulting process exhibits a quite different behaviour, a clear threshold can hardly be deﬁned in terms of the relevant dimensionless parameters, since Figs. 6.2b, c, d indicate that a smooth transition occurs from one behaviour to the other within a fairly narrow range of values of Shields parameter, say smaller than 0.1. This sensitive dependence of channel dynamics on the local value of Shields stress may turn out to pose a quite restrictive condition for the applicability of numerical models to predict the evolution of braided networks. We also note that, according to BT’s experiments with graded sediments and previous experimental and theoretical works (Lisle et al., 1991; Lanzoni and Tubino, 1999), sorting effects are likely to enhance the suppressing action of planform shape on bar migration and invariably lead to the formation of central wedge deposits of coarse particles acting as precursors of the bifurcation. Measured angles of streamlines of the two branches at bifurcation for both uniform and graded sediment experiments of BT are reported in Fig. 6.2e. Similar values of bifurcation angles have been observed by Zolezzi et al. (2006) and have been also reported by Federici and Paola (2003), who investigated the occurrence of bifurcations within a ﬂume consisting of a straight constant-width channel, with ﬁxed walls, followed by a linearly diverging reach, ending in a much wider constant width channel. Further information concerning the onset of the bifurcation has been obtained from the analysis of bank proﬁles. As pointed out before, the formation of bars leads to the establishment of a regular sequence of erosional bumps, whose amplitude increases as the planimetric forcing and channel widening inhibit further migration of bars. BT have found that the Fourier spectrum of bank proﬁles of channels evolving to a bifurcated state is always dominated by a component whose length coincides with that of bars and whose amplitude increases in time, until a maximum value is invariably reached at the onset of the bifurcation. In fact, once a bifurcating ﬂow establishes, it induces the erosion of bank regions that were previously undisturbed, which implies a reduction of the amplitude of bank oscillations, as shown in Fig. 6.2f. The occurrence of this maximum may provide a suitable criterion to set the onset of the bifurcation in terms of the aspect ratio and the Shields stress of the incoming ﬂow.

3.

Laboratory investigation on the equilibrium conﬁgurations of a bifurcation

We now turn to the relevant problem of deﬁning the equilibrium conﬁgurations of a bifurcation. The subject poses a number of questions that must be addressed. As

Bifurcations in gravel-bed streams

141

pointed out in ‘‘Introduction’’, channel bifurcations in gravel-bed braided rivers are almost invariably asymmetrical. Is this asymmetry related to the ﬁnal conﬁguration towards which the system is driven in its evolution? How long does the system take to attain an equilibrium state? Are these states stable or do channel bifurcations invariably evolve until one of the downstream branches eventually closes? Which is the inﬂuence exerted on the above conﬁgurations and on their stability by bed and channel processes resulting from the continuous rearrangement of the network? Furthermore, a bifurcation may occur as the result of a recently opened small branch, through chute cutoff or avulsion processes, or may initially consist of two channels receiving nearly balanced discharges as in anastomosed systems. Does the equilibrium conﬁguration depend on the initial state of the system? Answering the above questions preliminarily requires the collection of detailed experimental data along with the availability of a sound model able to reproduce the complex ﬂow structure which establishes at the bifurcation. Such a complexity has been recently highlighted by the numerical results of de Heer and Mosselman (2004), who tried to reproduce the experimental observations performed by Bulle (1926) in alluvial diversions with ﬁxed side-walls. The above results seem to indicate that the peculiar 3D structure of the ﬂow ﬁeld at the node may be responsible for a strongly unbalanced sediment partition in downstream branches, though the above effect is presumably much less intense in gravel-bed bifurcations which generally display relatively small ﬂow depths. In order to gain a deeper insight into the dynamical behaviour of a bifurcation we ﬁrst consider the results of experiments which have been recently performed with reference to a simple conﬁguration consisting of a symmetrical Y-shaped bifurcation, with ﬁxed walls, in which a main channel divides into two downstream distributaries. The experiments have been mainly devoted to measure water and sediment partition in the downstream branches, for different hydraulic conditions of the incoming ﬂow, and to provide a detailed characterization of ﬁnal bed topography attained by the system at equilibrium. The experimental runs have been carried out in the ‘p ﬂume’, a large experimental facility (25 m long and 3.14 m wide) located in the Hydraulic Laboratory of Trento University, which has been originally designed for scale models of braided networks. The ﬂume has been ﬁlled with well-sorted, sieved quartz sand, with mean diameter of 0.63 mm; a symmetrical bifurcation has been built inside the ﬂume, joining three channels with ﬁxed walls, rectangular cross section and movable bed. The widths of the upstream channel (a) and of downstream branches (b and c) have been set equal to 0.36 and 0.24 m, respectively, following the indication of rational regime theories which predict a non-linear relationship between the ﬂow discharge and the channel width, implying a ratio nearly equal to 1.3 between the total width of downstream channels and the upstream width, for a symmetrical conﬁguration (Ashmore, 2001). The bifurcation angle has been set to 301, with downstream branches diverging symmetrically from the direction of the upstream ﬂow. Further details on the experimental setting and procedures can be found elsewhere (Bertoldi et al., 2005; Bertoldi and Tubino, 2006). In order to determine the equilibrium conﬁguration of the bifurcation and to assess the inﬂuence of bars developing in the upstream branch, two different sets of

M. Tubino, W. Bertoldi

142

experiments have been performed. In each set different hydraulic conditions have been tested changing the incoming water discharge Qa and keeping the same initial value of the longitudinal bed slope. The sediment input has been regulated according to the transport capacity of the incoming uniform ﬂow estimated through Parker (1990) formula. In the ﬁrst set of runs, the slope has been set to the value of 0.3%: in this case, due to the relatively low values of the aspect ratio ba, the formation of free bars in the upstream channel was inhibited. In the second series the slope has been set to 0.7%: in all runs the formation of migrating alternate bars has been invariably observed in the upstream branch, since ba always exceeded the threshold value predicted by bar theories (Colombini et al., 1987). In few runs bars also developed in downstream branches; however, their effect on bifurcation has been always found to be negligible. The experimental conditions and the values of the relevant dimensionless parameters of the incoming ﬂow, namely the aspect ratio ba and the Shields stress Wa, are reported in Table 6.1, where S is the mean longitudinal slope at the end of the run and Da is the mean ﬂow depth measured in channel a. In the same table, we also summarize the measured equilibrium values of the discharge ratio of downstream branches rQ ¼ Qc =Qb , where the subscript ‘b’ always denotes the main downstream channel. Experimental ﬁndings indicate that, in the absence of backwater effects, the response of the system is mainly determined by ﬂow and sediment transport conditions

Table 6.1. Experimental conditions and measured values of discharge ratio and inlet step at equilibrium in Y-shaped bifurcations with ﬁxed walls. Run

S

Q (l/s)

Da (m)

ba

Wa

rQ

DZ

F3-18 F3-20 F3-21 F3-23 F3-25 F3-29 F3-37 F3-45 F3-61 F7-06 F7-07 F7-08 F7-09 F7-10 F7-12 F7-13 F7-15 F7-17 F7-20 F7-24

0.0031 0.0026 0.0027 0.0031 0.0026 0.0031 0.0033 0.0037 0.0029 0.0065 0.0066 0.0077 0.0078 0.0067 0.0070 0.0076 0.0076 0.0068 0.0078 0.0072

1.8 2 2.1 2.3 2.5 2.9 3.7 4.5 6.1 0.6 0.7 0.8 0.9 1 1.2 1.3 1.5 1.7 2 2.4

0.0167 0.0189 0.0191 0.0194 0.0215 0.0223 0.0254 0.0277 0.0363 0.0068 0.0075 0.0077 0.0083 0.0093 0.0102 0.0105 0.0113 0.0126 0.0134 0.0153

10.77 9.55 9.45 9.26 8.38 8.08 7.09 6.50 4.95 26.30 23.91 23.30 21.66 19.38 17.71 17.21 15.88 14.28 13.45 11.73

0.0459 0.0425 0.0453 0.0524 0.0487 0.0599 0.0721 0.0873 0.0855 0.092 0.084 0.082 0.076 0.0574 0.0655 0.0734 0.0787 0.0784 0.0943 0.0985

0.294 0.466 0.56 0.730 0.65 0.8 0.91 0.99 0.97 0 0 0 0.25 0.05 0.45 0.5 0.5 0.45 1 1

0.694 0.594 0.514 0.365 0.312 0.153 0.043 0.022 0.019 1.307 1.370 1.726 1.318 1.228 0.821 0.773 0.397 0.722 0.485 0.198

Bifurcations in gravel-bed streams

143

in the upstream channel and follows two distinct behaviours. At relatively high values of Wa the bifurcation remains stable, keeping a balanced partition of water and sediment discharge in downstream branches (or returning to a balanced state once perturbed with an extra amount of sediment feed in one branch). At lower values of Wa, in spite of the symmetrical character of the bifurcation, the system evolves towards an unbalanced conﬁguration, where different ﬂow and sediment discharges are conveyed in downstream branches, such that rQ attains equilibrium values invariably lower than 1. Examples of the temporal evolution of both balanced and unbalanced bifurcations are given in Fig. 6.3a. In the latter case the system undergoes strong morphological changes, such that the ﬁnal bed conﬁguration is characterized by a distinctive asymmetry, which is found to replicate the main features displayed by natural bifurcations. In particular, local aggradation occurs at the node, while the average bed level reached by downstream branches is significantly different, as shown in Fig. 6.3b, the main downstream channel being invariably located at a lower level. Hence, a transverse inlet step establishes at the node, whose amplitude DZ, scaled with the ﬂow depth Da, is also reported in Table 6.1 (such parameter has been computed as the difference between the values of bed elevation at the inlet of distributaries obtained through linear interpolation of their longitudinal bed proﬁles). The uneven partition of ﬂow discharge at the node is sustained by the bed conﬁguration of the upstream channel, where a transverse deformation gradually establishes, in the form of a steady alternate bar, over a length of few channel widths upstream of the bifurcation. Such depositional bar, whose intensity depends on local ﬂow conditions, drives the ﬂow and the sediment transport towards the main branch; its occurrence reﬂects the upstream morphological inﬂuence exerted by the sudden deviation of the main ﬂow orientation induced by the bifurcation. Measured equilibrium values of the discharge ratio rQ and of the inlet step DZ are reported in Figs. 6.3c and d in terms of the Shields stress and of the aspect ratio of the upstream ﬂow, respectively, for the whole set of experiments. It is worth noticing that, for a given channel slope, increasing values of Wa correspond to decreasing values of ba. Hence, the stronger is the degree of asymmetry of bed conﬁguration, that is the larger is DZ, the smaller is the discharge ratio rQ, which implies a highly unbalanced bifurcation. The above results highlight the strong link between the discharge distribution and the bed topography at the bifurcation point, which is implied by the almost linear relationship between rQ and DZ. For those runs in which alternate bars did not occur such relationship takes the following simple form: DZ ¼ 1 rQ

(6.2)

Experimental results show that a Y-shaped bifurcation is driven towards more unbalanced equilibrium conﬁgurations as the Shields stress decreases. At low values of sediment mobility the system becomes unable to keep active both downstream branches, such that one branch closes and rQ vanishes. Under these conditions the measured difference of bed level at the inlet of downstream branches is comparable with the average depth of the incoming ﬂow. It is worth noting that the above results conform to those recently obtained by Federici and Paola (2003), who investigated

M. Tubino, W. Bertoldi a)

b)

1.2

0 -0.2 -0.4 -0.6 -0.8 -1 -1.2 -1.4 -1.6 -1.8 -2

bed elevation [cm]

discharge ratio rQ

144

1 0.8 0.6 0.4 0.2 0 00.00

run F3-21 run F3-45 04.00

c)

08.00 12.00 time[hours]

16.00

0

20.00

d)

0.6 0.4

50 100 150 200 250 longitudinal coordinate [cm]

1 0.8 0.6 0.4

0.2

0.2 0 0.04

e) 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0 0.00

0.06 0.08 Shields stress

0.1

0.12

a

0

5

f)

10 15 aspect ratio a

20

25

20.00

0.00

discharge ratio rQ

7

2.00

4.00 time [hours]

6.00

6 5 4 3 2 1 0 0.00

8.00

4.00

8.00 12.00 16.00 time [hours]

h) 1.4

g) 1

1.2 inlet step ∆η

discharge ratio rQ

S = 0.3% S = 0.7%

0.8 0.6 0.4 0.2

1 0.8 0.6 0.4 S = 0.3% S = 0.7%

0.2 0

0 -1

300

S = 0.3% S = 0.7%

1.2

0.8

0 0.02

discharge ratio rQ

1.4

S = 0.3% S = 0.7% inlet step ∆

discharge ratio rQ

1.2 1

right channel left channel

0

1 2 (a-R)/R

3

4

-1

0

1 2 (a-R)/R

3

4

Figure 6.3. Experiments in Y-shaped bifurcations with ﬁxed banks: (a) time evolution of the discharge ratio rQ measured in the balanced run F3-45 and in the unbalanced run F3-21; (b) longitudinal bed proﬁles of the downstream branches (run F3-21); (c) and (d) time evolution of the discharge ratio measured in two runs in which alternate bars have been observed (runs F7-24 and F7-08); (e) discharge ratio and (f) inlet step at equilibrium as a function of the Shields stress and the aspect ratio of the upstream channel; (g) discharge ratio and (h) inlet step at equilibrium as a function of the distance from resonant conditions.

Bifurcations in gravel-bed streams

145

the stability of bifurcations occurring in a diverging channel, fed by an upstream channel with constant width and ﬁxed walls. They observed that the bifurcation was stable and fairly symmetrical for relatively large values of the Shields parameter of the incoming ﬂow; at lower values, say smaller than 0.15, bifurcations were unstable since the ﬂow switched repeatedly and randomly from one branch to the other. Figs. 6.3c and d also suggest that the development of free alternate bars in the upstream channel, as observed in the second series of experiments, does not seem to change the above scenario, at least qualitatively (the corresponding equilibrium values of rQ and DZ are reported as closed triangles). However, we have found that bar occurrence enhances the tendency of the system towards the establishment of unbalanced conﬁgurations, since in this set of experiments we have invariably measured smaller values of the discharge ratio, and consequently larger values of the inlet step, for similar values of the dimensionless parameters of the incoming ﬂow. Furthermore, a more complex and irregular evolutionary process has been detected, such that the system has never achieved a steady conﬁguration, unless in the case of strongly unbalanced bifurcations. Examples of the measured variations of the discharge ratio in the experiments characterized by the presence of migrating bars are reported in Figs. 6.3e and f both for a stable and an unstable bifurcation. In the former case, the discharge distribution remains almost balanced: the effect of migrating bars is reﬂected by the ﬂuctuations of rQ around the equilibrium value. In the latter case, which corresponds to smaller values of Shields stress, the presence of migrating alternate bars causes, since the beginning of the experiment, a sudden instability of the bifurcation, which leads to a strongly unbalanced conﬁguration. Due to the morphodynamic inﬂuence exerted by the bifurcation, a steady bed deformation establishes in the incoming channel, whose amplitude is comparable with that of migrating bars, which in turn affects bar migration. However, bedforms migration is still able to destabilize the system such that the ﬂow switches from one channel to the other. The above behaviour highlights the crucial role played by the mutual interaction between steady and migrating bed responses on the stability of the bifurcation. It also suggests the opportunity to revisit the experimental ﬁndings in the light of the concept of morphodynamic inﬂuence, as it has been stated theoretically by Zolezzi and Seminara (2001) and more recently conﬁrmed experimentally by Zolezzi et al. (2005). The experimental runs summarized in Table 6.1 correspond to both sub-resonant and super-resonant conditions. If we report the measured values of rQ and DZ in terms of the relative distance from the resonant conditions ðb bR Þ=bR , as in Figs. 6.3g and h, we ﬁnd that experimental data fall approximately on the same curve: sub-resonant conditions typically correspond to balanced equilibrium conﬁgurations, while in super-resonant conditions the bifurcation evolves towards unbalanced conﬁgurations. Furthermore, increasing the distance from the resonant value bR, the degree of asymmetry of the bifurcation becomes larger. Such results can be given a relatively simple physical explanation. In super-resonant conditions the morphodynamic inﬂuence of the bifurcation is mainly felt upstream, which implies that, as the main ﬂow changes its orientation at the node, a transverse bed deformation occurs in the upstream branch, which in turn determines an unbalanced

M. Tubino, W. Bertoldi

146

ﬂow partition. In sub-resonant conditions the inﬂuence of the node can be only reﬂected by downstream topography. Hence, the bifurcation can keep stable. These results, which need to be conﬁrmed by further experimental evidence, seems to suggest the possibility of interpreting the response of a bifurcation to the variations of the incoming ﬂow in terms of a single parameter which measures the distance of local ﬂow conditions from the resonant range.

4.

The theoretical model of Bolla Pittaluga et al. (2003)

Theoretical models which have been proposed so far to investigate the stability of channel bifurcations essentially differ for the different structure of the nodal point conditions accounting for the sediment exchange at the node (Slingerland and Smith, 2004). A proper formulation of such conditions has been found to be the crucial ingredient, which is required to replicate the observed characteristics of natural bifurcations. For gravel-bed bifurcations a relatively simple model has been recently proposed by Bolla Pittaluga et al. (2003) (hereinafter referred to as BRT) within the context of a 1D approach. The model considers the same conﬁguration analyzed in the previous section, namely a Y-shaped bifurcation with ﬁxed banks; however, the possible development of migrating bars in the incoming channel is not accounted for. BRT’s approach follows closely the procedure originally introduced by Wang et al. (1995), according to which the equilibrium conﬁgurations of a Y-shaped bifurcation are determined through the solution of ﬁve nodal point conditions, which must be imposed at the node in a 1D model, along with two stage discharge relationships for the downstream branches. Using the same notation adopted in the previous section, the above conditions read: Qa ¼ Q b þ Q c ; Hi ¼ Ha Q i ¼ bi C i D i

(6.3)

ði ¼ b; cÞ; pﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ gRi Si

(6.4a,b) ði ¼ b; cÞ.

(6.5a,b)

They impose the balance of water discharge Q and the constancy of water level H at the node, along with uniform ﬂow conditions in downstream branches, where C is Chezy resistance coefﬁcient and R the hydraulic radius. Wang et al. (1995) have used an empirical condition to model the partition of sediment discharge at the node, along with a condition similar to (6.3) expressing the sediment discharge balance. An alternative approach has been proposed by BRT, which is able to account for topographically driven effects on ﬂow and sediment transport which occur just upstream of the bifurcation. They have employed a ‘‘quasi two-dimensional approach’’ close to the bifurcation, whereby the last reach of the upstream channel, over a length of few channel widths, aba , has been divided in two adjacent cells, both fed with a sediment discharge proportional to their upstream width and each feeding a downstream branch. Applying Exner (1925) equation to both cells they have derived the following nodal point conditions for sediment

Bifurcations in gravel-bed streams

147

transport: 1 ba dZi qi qa ðba =ðbb þ bc ÞÞ qy þ ¼0 ð1 pÞ 1 þ bb þ bc dt bi 2 aba

ði ¼ b; cÞ,

(6.6a,b)

where t is time, p the sediment porosity, Z the bed elevation, q the sediment discharge per unit width; furthermore, qy the transverse exchange of sediments between the two cells due to lateral ﬂow exchange and gravitational effects, which has been evaluated through a suitable generalization of the relationship used to describe bedload transport over a gently sloping bed in quasi-unidirectional ﬂow (e.g., Ikeda et al., 1981). We note that qy depends both on the discharge ratio rQ and on the inlet step DZ. Hence, this term embodies the fundamental mechanism retained in BRT model: whenever one of the downstream branches receives less water or its average bed level is higher than the level of the other branch, the resulting excess of bedload is diverted to the other branch. Due to the inclusion of this effect, the model of BRT is able to capture the main distinctive features of gravel-bed bifurcations. In particular, according to model results, the behaviour of the bifurcation is found to depend upon the values of relevant parameters describing the incoming ﬂow, namely Wa, ba and Sa. In agreement with experimental ﬁndings for a symmetrical bifurcation (bb ¼ bc ), stable equilibrium solutions in which the two downstream branches are fed with the same water and sediment discharges are only possible for relatively large values of the Shields parameter Wa; for smaller values of Wa, say smaller than 0:1C0:15, a further equilibrium solution appears (along with its reciprocal), which is found to be invariably stable and implies a strong imbalance of discharge partition in downstream branches. The threshold curves below which the balanced solution is no longer stable, as obtained by BRT, are plotted on the plane (Wa, ba) in Fig. 6.4a, for different values of the upstream slope Sa. As shown in Fig. 6.4b the agreement between theoretical predictions and experimental ﬁndings is fairly good, particularly for the set of experiments in which bars didn’t form (in Fig. 6.4b values of the discharge ratio rQ larger than 0.9 have been considered as implying balanced bifurcations). The theoretical model also reproduces the observed asymmetry of bed conﬁguration at the node, though with an accuracy compatible with its simple 1D character. In fact, unbalanced solutions are found to be invariably characterized by different values of the ﬂow depth in downstream branches; through equation (6.4) this implies that the local bed levels differ at the node such that an inlet step forms. Furthermore, the model replicates the observed tendency towards the establishment of strongly unbalanced conﬁgurations for smaller values of Wa (or larger values of ba), and predicts the corresponding increase of the amplitude of the inlet step. Stability diagrams for a Y-shaped symmetrical bifurcation are reported in Figs. 6.4c and d. They are given in terms of the equilibrium values of the discharge ratio rQ ¼ Qc =Qb and of the amplitude of inlet step DZ. Stable solutions are denoted by solid lines, while dashed lines indicate unstable solutions. We note that unlike Fig. 6.9 of BRT, for which Meyer-Peter and Muller’s transport relationship was employed, Figs. 6.4c and d have been obtained using Parker (1990)’s bedload transport relationship, which behaves smoothly as the Shields stress tends to zero.

M. Tubino, W. Bertoldi

148 b)

0.5 0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0

0.12

BALANCED SOLUTION UNSTABLE

0.08 0.06 0.04 0.02

0

5

10

15

20

0

25

aspect ratio a

5

10

15

20

25

30

aspect ratio a

d)

10 discharge ratio rQ

unbalanced runs balanced runs

0

c)

1

inlet step ∆

0.5

1

0

-0.5

0.1

-1 0

5

e)

10 aspect ratio a

15

20

0

5

f)

100

10 aspect ratio a

15

20

1

10

100

1.2 0.8

10

inlet step ∆

discharge ratio rQ

Bolla Pittaluga et al. (2003)

0.1

a

S = 0.005 S = 0.01 BALANCED S = 0.025 SOLUTION STABLE

Shields stress

Shields stress

a

a)

1 0.1

0.4 0 -0.4 -0.8

0.01 0.01

0.1

1 slope ratio rS

10

100

-1.2 0.01

0.1

slope ratio rS

Figure 6.4. Theoretical results of Bolla Pittaluga et al. (2003): (a) equilibrium conﬁgurations of a symmetrical Y-shaped bifurcation (the threshold line separates the region where the balanced conﬁguration is the only stable solution from the region where two further unbalanced solutions occur); (b) comparison between the experimental results of Bertoldi and Tubino (2006) and the theoretical predictions; (c) discharge ratio and (d) inlet step as a function of the aspect ratio of the upstream channel (W ¼ 0:1; S ¼ 0:007); (e) discharge ratio and (f) inlet step as a function of the slope ratio of the distributaries (b ¼ 15; W ¼ 0:1; S ¼ 0:007).

It is worth noting that the governing mathematical problem, set in terms of the equations (6.3)–(6.6), displays a further degree of freedom, in that the stability of the bifurcation is also affected by the downstream boundary conditions through the dependence on downstream slopes included in (6.5). In particular, the stability diagrams reported in Figs. 6.4c and d refer to equilibrium solutions for which the values of downstream slopes coincide, namely rS ¼ S c =S b ¼ 1: otherwise, balanced solutions cannot occur. In the work of BRT this effect has been included through the dependence of the solution on the ratio of the lengths of downstream channels (see Fig. 6.7 of BRT). Implicit in their procedure is the assumption that the downstream conditions set a unique value of the water surface elevation for both branches, which

Bifurcations in gravel-bed streams

149

may represent the case of a closed loop where channels rejoin at a downstream node as well as that of two branches delivering into a pond. In this respect, the model of BRT follows the classical viewpoint of long-term morphodynamic analysis. In fact, following the lead of Wang et al. (1995), BRT assume that the time evolution of the channel loop can be described as a sequence of uniform ﬂow conditions, such that sediment continuity is applied to downstream branches in global form, accounting for the sediment discharge entering each channel through the upstream mouth and leaving it from the outlet section. One can readily argue that the above approach implies that the stability of the system is investigated on a relatively long timescale, namely that required for the morphological bed response of channels joining at the node to adapt to ﬂow discharge and to the effect of downstream boundary conditions. However, due to the relatively small values of Shields parameter typical of gravel-bed channels, the timescale required for the morphodynamic inﬂuence of downstream boundary conditions to be felt may be quite large when compared with the timescale over which significant morphological changes occur at the node, like those which lead to node shifting or annihilation. An alternative approach has been recently proposed by Miori et al. (2006) (see also Hirose et al., 2003), whereby a local analysis has been introduced which neglects the effect of downstream boundary conditions and assumes that the ﬂow in downstream branches is in equilibrium with the local bed slope, according to a suitable stage–discharge relationship. The above approach is obviously unable to account for backwater effects, which may be induced, at high ﬂows, by the physical constraint imposed by the geometry of slopes ﬂanking the braided reach or by the presence of a regulating weir at the outlet, as in the case of Ridanna Creek (Zolezzi et al., 2006). However, a local analysis seems more suitable to describe the inherent response and the stability of bifurcations subject to relatively rapid changes, as it occurs in gravelbed braided rivers. It is worth noting that the adoption of the latter viewpoint does not affect the resulting pattern of equilibrium conﬁgurations and their stability; however, it leads to a different definition of the timescale of the evolutionary process, which loses it dependence on the length of downstream channels. Hence, the above distinction may turn out to be relevant for the assessment of the relative role on the stability of the bifurcation of further processes, like channel width adjustment or bar migration, each of them characterized by a distinctive timescale. The dependency on the slope ratio rS ¼ S c =Sb of the equilibrium values of the discharge ratio and of the amplitude of inlet step, as predicted by the theoretical model, is shown in Figs. 6.4e and f. It appears that for values of rS falling within a convenient range around rS ¼ 1 three conﬁgurations are possible, where the ﬂow discharge is always unevenly apportioned between branches, unless rS ¼ 1. Furthermore the solution corresponding to the less unbalanced conﬁguration (dashed lines in Figs. 6.4e and f) is always unstable. Hence, within this range of values of rS the bifurcation invariably evolves towards a highly unbalanced conﬁguration and the ﬂow switch to a given branch is not conditioned by the local values of channel slope. For values of the slope ratio rS falling outside the above region, the systems only

M. Tubino, W. Bertoldi

150

admits of one solution, which is strongly unbalanced and conveys the ﬂow in the channel characterized by the larger slope.

5.

Channel bifurcations with erodible banks

Setting the banks of the channels to be ﬁxed, as we have done in the preceding sections, is equivalent to consider that the timescale of planimetrical changes is much larger than the timescale of bed evolution. This is a quite common assumption for single-thread meandering channels, where bank erosion is mainly controlled by sediment cohesion and vegetation. In gravel-bed braided rivers, however, the timescales of bank and bed erosion may be comparable and single channels can be often considered, up to a certain extent, as laterally unconstrained (Murray and Paola, 1994). Field evidence suggests that during formative events the overall structure of the braided network undergoes major changes. It is then reasonable to expect that, as the bifurcation evolves, the channels joining at the node can adjust their width to the ﬂowing discharge. Whether or not channel width adaptation can proceed such as to accomplish a regime relationship with the actual ﬂow conditions crucially depends on the timescale of the process, as compared with the typical timescale of the bifurcation evolution. We also note that, if we leave the channel width to be free to change, a further element of asymmetry is introduced in the process. In fact a gravel-bed channel widens when its discharge increases. On the contrary, the counterpart process of channel narrowing does not occur in response to a discharge fall, at least on a relatively short timescale. The above considerations have motivated two further series of experiments that have been carried out in the ‘p ﬂume’. In the ﬁrst set of experiments (‘‘M’’-type) we have removed the ﬁxed banks of downstream branches, leaving the upstream channel at the original width of 0.36 cm. In the second set (‘‘L’’-type) the channels joining at the node have been traced directly on the sand-plain, leaving the whole system to be free to evolve. Hence, while in the former series the incoming conditions were ﬁxed and the observed evolution has been mainly induced by the variation of the shape, orientation and position of the node, as well as by downstream channels widening, in the latter series the incoming conditions changed during each run and the evolutionary processes of the bifurcation have been also inﬂuenced by the slow planimetric evolution of the weakly meandering upstream branch. Experimental conditions are summarized in Table 6.2. We may note that in all cases the system evolved towards unbalanced conﬁgurations. However, measured data do not display a regular trend as in the case of ﬁxed-bank channels. This reﬂects the highly dynamical character of these runs, whose evolution has been markedly conditioned by the interaction of various bed and channel processes, which often prevented the achievement of stable equilibrium conﬁgurations (measured data reported in the table correspond to quasi-equilibrium conditions which lasted for a sufﬁciently long time span). Notwithstanding such complexity, the above experiments provide a consistent picture of some distinctive features of the evolutionary process of a bifurcation. The

Bifurcations in gravel-bed streams

151

Table 6.2. Experimental conditions and measured values of discharge ratio and inlet step at quasi-equilibrium in bifurcations with erodible banks. Run

S

Q (l/s)

Da (m)

ba

Wa

rQ

DZ

M3-19 M3-21 M3-23 M3-25 M3-28 M3-33 L7-08 L7-10 L7-12 L7-15

0.0027 0.0032 0.003 0.0029 0.0028 0.0033 0.0063 0.0062 0.0066 0.006

1.9 2.1 2.3 2.5 2.8 3.3 0.8 1 1.2 1.5

0.0179 0.0182 0.0195 0.0209 0.0225 0.0236 0.0155 0.0141 0.0151 0.0162

10.1 9.9 9.2 8.6 8 7.6 11.1 14.4 15.9 16.8

0.0431 0.0511 0.0516 0.0522 0.0549 0.0678 0.0771 0.0707 0.0803 0.0802

0.14 0.22 0.02 0.2 0.04 0.14 0.7 0.6 0.4 0.55

0.546 0.464 0.819 0.529 0.725 0.554 0.296 0.128 0.378 0.164

experimental ﬁndings suggest that the time behaviour of the discharge ratio rQ observed in most of the runs can be approximated through an exponentially decreasing function towards the equilibrium value, as shown in Fig. 6.5a. Hence, from these data a suitable timescale can be determined, through a best-ﬁt procedure, which may be taken as a representative scale of the dynamical response of the bifurcation. In order to ﬁlter out the dependence of the above scale on bedload intensity, which measures the speed of any morphological process obeying to Exner (1925) equation, we have scaled the time variable using the morphological timescale bD t ¼ qﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ , gDD3s F

(6.7)

where D is the sediment relative density and F the bedload intensity. Interpolated values of the dimensionless timescale T are plotted in Fig. 6.5b in terms of the parameter ðba bR Þ=bR , which measures the relative distance from the resonant conditions in the upstream channel. Data refer both to ‘‘F’’-type experiments (ﬁxed banks) and to ‘‘M’’-type experiments. In the latter runs we have invariably observed a slower evolution, as a consequence of the adjustment process of the width of downstream branches. Hence, as shown in Fig. 6.5b the resulting timescale T is larger. Experimental ﬁndings also suggest that, in terms of the scaled time variable, the evolutionary process is much faster under super-resonant conditions, when the equilibrium conﬁguration is more unbalanced, while T increases sharply as we cross the resonant range. We note that T provides a measure of the ratio between the timescale of the bifurcation and the morphological scale (6.7) and does not represent the actual speed of the process. Within the super-resonant range typical values of T are much smaller than 1, which implies that the bifurcation evolves on a relatively fast timescale. All the experiments performed with erodible banks have displayed a quite similar behaviour. A distinctive sequence of phases has been identiﬁed, which reﬂected the inﬂuence exerted by various morphological processes acting on different timescales. Such behaviour can be reconstructed on the basis of the results reported

M. Tubino, W. Bertoldi

152 a)

b)

discharge ratio rQ

1.2

0.8 0.6 0.4 0.2 0 0

c)

0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 dimensionless time t*

discharge asymmetry AQ

1 0.8 0.6 0.4 0.2 0 0.00 -0.2 -0.4 -0.6

2.00

4.00

6.00

period of rQ oscillations [min]

time scale T

1

1.60 1.40 1.20 1.00 0.80 0.60 0.40 0.20 0.00 -0.4

Fruns Mruns

-0.2

d)

0

0.2 0.4 (a-R)/R

0.6

0.8

1

80 70 60 50 40 30 20 10 0 0

time [hours]

10

20 30 40 50 60 70 period of bar migration [min]

80

migration speed [m/s]

e) 0.0018 0.0016 0.0014 0.0012 0.001 0.0008 0.0006 0.0004 0.0002 0 0.00

0.15

0.30

0.45

1.00

1.15

1.30

time [hours]

Figure 6.5. (a) The time evolution of the discharge ratio measured in run M3-19 and the corresponding exponential curve; (b) the dimensionless timescale T as a function of the relative distance from resonant conditions; (c) time evolution of discharge asymmetry measured in the experimental run L7-10; (d) relationship between the period of oscillation of the discharge ratio and the period of bar migration; (e) damping of bar migration speed observed in run L7-10.

in Figs. 6.5c, d, e. An example of the observed time evolution of discharge asymmetry AQ ¼

Qb Qc Qb þ Qc

(6.8)

is given in Fig. 6.5c. In the ﬁrst phase AQ ﬂuctuates due to the development of bars in the incoming channel, which eventually reach the inlet of downstream branches such that ﬂow switches its direction; at this stage the period of oscillation is ﬁxed by the migration speed of bars, as shown in Fig. 6.5d. In the meantime the node shifts downstream, due to the high erosion rate of incoming ﬂow, the angle between downstream branches increases, reaching a maximum value ranging 60–701, and a steady bed pattern gradually establishes just upstream of the node. Furthermore, the channels joining at the node widen and modify their planimetric conﬁguration, assuming a weakly meandering pattern.

Bifurcations in gravel-bed streams

153

These processes affect in turn the development of bars, and lead to a marked and rapid decrease of their speed, as shown in Fig. 6.5e, while their length remains almost ﬁxed in spite of channel widening, like in BT experiments. Furthermore, they lead to the establishment of a dominant channel and discharge asymmetry AQ rapidly increases until it reaches a peak value. This also drives the incoming bars towards the main downstream branch. However, as two consecutive bars enter in the same channel, the ﬂow suddenly switches to the other branch. The subsequent evolution of the bifurcation is mainly determined by the planimetric evolution of the channels joining at the node. A further switch of the conﬁguration is determined on a longer timescale by the planimetric shift of the weakly meandering upstream branch, though small-scale ﬂuctuations due to migrating bedforms are still detectable. The bifurcation angle decreases and eventually reaches a value falling in the range 45–601, as observed in other laboratory experiments (see Section 2). The above observations provide a picture which is far from being conclusive and would need to be integrated with further detailed measurements on the behaviour of bifurcations within a network, such as to account for the effect brought by the interaction between the various objects constituting the network and by the preceding development of the network itself. However, experimental ﬁndings unequivocally suggest that balanced bifurcations can hardly be expected. Furthermore, they show that the evolution of the node is strongly related both to width changes and to channel shift. A recent attempt to include the former effect in the model of BRT is due to Miori et al. (2006) (hereinafter referred to as MRT). In the analysis of BRT the channel widths are ﬁxed and are prescribed as input data. In order to account for channel width adaptation a suitable relationship is needed to relate the channel width to the hydraulic characteristics. However, as pointed out by Chew and Ashmore (2001), empirical formulas based on data collected in single-thread channels often fail in describing the longitudinal width changes of braided reaches. Water discharge indeed dominates empirical regime relations, while ﬁeld data collected on a braided reach of the Sunwapta River suggest that the effect of grain size plays an important role on channel width adjustment. The above limit is partially removed by rational regime formulas which, however, still predict rather unprecise values of the observed channel width, despite accounting for the effect of grain size. In their analysis MRT have adopted both a rational regime formula and an empirical relationship, namely: b ¼ 5:28QS 1:26 D3=2 s b ¼ 0:087O0:599 D0:445 s

ðGriffiths; 1981 Þ ðAshmore; 2001Þ

(6.9)

(6.10)

where O ¼ gSQ is the stream power and g the water speciﬁc weight. We note that (6.10) is an empirical formula specifically designed for gravel-bed channels, which has been proposed by Ashmore (2001) on the basis of a statistical regression on ﬁeld data.

M. Tubino, W. Bertoldi

154

a)

b)

0.05

0.08 0.07 0.06 0.05 0.04 0.03 0.02 0.01 0

Shields stress ϑ

Shields stress

Two novelties are introduced with respect to BRT in the analysis of the equilibrium conﬁgurations of a Y-shaped bifurcation. A ﬁrst novelty is brought in the dimensionless representation of the incoming ﬂow, as the adoption of (6.9) or (6.10) imposes a given relationship among the dimensionless parameters of the upstream ﬂow Wa, Sa and ba, as plotted in Figs. 6.6a and b. In particular we note that the empirical formula (6.10) implies a slight increase of Shields stress as the aspect ratio increases, which is also implied by any rational regime formula, analogous to (6.9), which although displays a much weaker dependence. Furthermore, for any given set of dimensionless parameters characterizing the incoming ﬂow, the governing system (6.3)–(6.6) must be supplemented with two further relationships of the form (6.9) or (6.10) to account for width adjustment of downstream branches to the local ﬂow conditions. MRT have found that in both cases, whatever value of the aspect ratio is used, the Y-shaped conﬁguration

0.04

0.03

S = 0.003 S = 0.007 S = 0.015

0.02 0

5

c)

10 15 aspect ratio

20

25

0

10 15 aspect ratio

20

25

10 width ratio rb

discharge ratio rQ

5

d)

100 10 1 0.1

1

0.1

0.01 0

5

e)

10 15 aspect ratio a

20

25

0

5

0

1

f)

1

1

0.8

0.8

width ratio rb

discharge ratio rQ

S = 0.003 S = 0.007 S = 0.015

0.6 0.4 0.2 0

10 15 aspect ratio a

20

25

2 3 4 dimensonless time t /

5

0.6 0.4 0.2 0

0

1

2 3 4 dimensonless time t /

5

Figure 6.6. Theoretical results of Miori et al. (2006): the relationship between Shields parameter and aspect ratio, for different values of channel slope, as obtained through Grifﬁths (1981) formula (a) and Ashmore (2001) relationship (b); the equilibrium values of rQ (c) and of DZ (d) reached by the system starting from a symmetrical initial condition, as a function of ba (Sa ¼ 0:007); the time evolution of the discharge ratio rQ (e) and of the width ratio rb (f) starting from different initial conditions (Sa ¼ 0.007).

Bifurcations in gravel-bed streams

155

invariably admits of unbalanced equilibrium solutions. This is shown in Fig. 6.6c where we report the equilibrium states of the bifurcation, in terms of the equilibrium values of discharge ratio rQ, as obtained using the empirical relationship (6.10) and the transport formula proposed by Parker (1990), for rS ¼ Sc =S b ¼ 1. We note that the resulting inability of the system to maintain a balanced conﬁguration when banks are erodible conforms to the results of experimental observations. The model of MRT also provides an estimate of the equilibrium width of downstream channels: as shown in Fig. 6.6d, for typical values of the aspect ratio of the incoming ﬂow, the main downstream channel at equilibrium may be 2–4 times wider than the smaller channel. In order to investigate the stability of such equilibrium conﬁgurations MRT have adopted a local approach. As discussed before, this implies that the time evolution of the system is no longer dependent on the length of downstream branches, as in BRT. As a consequence, the speed of the process is governed by the nodal conditions (6.6a, b) and, through the remaining set of nodal equations, is related to the local characteristics of the bifurcation. Furthermore, in order to account for the different response of the channel width to discharge rise and fall, MRT have introduced a simpliﬁed approach such that channel widening has been computed through the empirical formula of (6.10), while channel narrowing has been precluded. This simple mechanism turns out to be quite important to disclose a further property of the system. In fact, MRT analysis reveals that, in addition to the previously discussed equilibrium conﬁguration, which essentially requires that both downstream channels conform their width to a regime formula, further equilibrium states may exist which depend on the initial state. In fact, according to the model of MRT the bifurcation cannot reach a ﬁnal conﬁguration in which one branch is narrower than its initial value. It is worth noting that the initial condition, which can be arbitrarily imposed in the model, essentially embodies the effect of different inception mechanisms. Hence, the main consequence of the results of MRT is that the equilibrium conﬁgurations are found to depend on the mechanism through which the bifurcation is formed. Results of MRT are summarized in Figs. 6.6e and f, where we plot the time evolution of the discharge ratio rQ ¼ Qc =Qb and of width ratio rb ¼ bc =bb of a Y-shaped bifurcation for different initial values of the width ratio rb of downstream branches. The equilibrium value of the above ratios corresponding to regime conditions of both downstream branches are reported with dotted lines. Fig. 6.6f suggests that the geometry of the bifurcation, namely the ratio between the widths of the downstream distributaries, is quite sensitive to the initial conditions, such that the possible equilibrium states covers almost the whole range of values. On the contrary, Fig. 6.6e shows that the range of equilibrium values of rQ is quite narrow. Furthermore, when the initial value of rb is larger than the corresponding equilibrium value under regime conditions, the ﬁnal conﬁguration attained by the system is strongly reminiscent of the initial state. This may be the case of distributive bifurcations, where the initial channels display almost comparable widths. On the other hand, a bifurcation originating from the incision of an initially narrow branch leaving the main channel is driven towards an equilibrium

M. Tubino, W. Bertoldi

156

conﬁguration which does not differ appreciably from that occurring when both downstream branches satisfy a regime relationship. In fact, in this case the bifurcation is initially characterized by a strong imbalance, which implies that the subsequent evolution of the bifurcation is dominated by widening of the smaller channel, while the width of the main branch remains almost ﬁxed.

6.

Concluding remarks

Recent theoretical and experimental investigations on gravel-bed rivers bifurcations provide a deeper insight into one of the dynamical processes that mainly controls the complexity of braided networks. The recurring asymmetrical conﬁgurations observed in the ﬁeld are reproduced both by laboratory models and by simpliﬁed theoretical models, which indicate an intrinsic tendency towards unbalanced conﬁgurations mainly due to local effects. This picture does not change significantly due to external controls, as those exerted by bar occurrence and by channel shift. It is also shown how equilibrium in cohesionless bifurcations with erodible banks is strongly sensitive to the initial conﬁguration: such dependence is quite relevant for natural streams subject to irregular duration and intensity of the forming events. The above results, far from being exhaustive, may significantly enhance the chances of success in predicting channel pattern evolution of braided streams, provided they can be transformed in simpliﬁed rules to be incorporated into predictive numerical models (e.g., Paola 2001, Jagers 2003). Research on this topic would beneﬁt from a deeper investigation of the interplay between the intrinsic timescale of bifurcation and those related to migrating bars and determined by the duration of formative events. Finally, a more complete understanding of bifurcation dynamics within a braided network will require further insight into the complex interaction of local processes with channel and node shifting occurring at various locations in the braided river.

Acknowledgments This work has been developed within the framework of the ‘Centro di Eccellenza Universitario per la Difesa Idrogeologica dell’Ambiente Montano – CUDAM’, the projects ‘Morfodinamica delle reti ﬂuviali – COFIN2001’ and ‘La risposta morfodinamica di sistemi ﬂuviali a variazioni di parametri ambientali – COFIN 2003’, co-funded by the Italian Ministry of University and Scientiﬁc Research (MIUR) and the University of Trento, and the project ‘Rischio Idraulico e Morfodinamica Fluviale’ ﬁnanced by the Fondazione Cassa di Risparmio di Verona, Vicenza, Belluno e Ancona. The authors gratefully acknowledge Guido Zolezzi for carefully reading the paper and Stefano Miori for the plots. The authors are deeply thankful to all the people who joined and participated in the laboratory activities and data processing, in particular to Stefania Baldo, Alessio Pasetto and Luca Zanoni.

Bifurcations in gravel-bed streams

157

References Ashmore, P., 1991. How do gravel-bed rivers braid? Can. J. Earth Sci. 28, 326–341. Ashmore, P., 1988. Bed load transport in braided gravel-bed stream models. Earth Surf. Process. Landf. 13, 677–695. Ashmore, P., 2001. Braiding phenomena: statics and kinetics. In: Mosley, M.P. (Ed.), Gravel-Bed Rivers V. New Zealand Hydrological Society, Wellington, pp. 95–121. Ashworth, P., 1996. Mid-channel bar growth and its relationship to local ﬂow strength and direction. Earth Surf. Process. Landf. 21, 103–123. Bertoldi, W., Pasetto, A., Zanoni, L., Tubino, M., 2005. Experimental observations on channel bifurcations evolving to an equilibrium state. In: Proceedings of RCEM2005 Conference, Urbana, IL, USA, 4–7 October, pp. 409–419. Bertoldi, W., Tubino, M., 2005. Bed and bank evolution of bifurcating channels. Water Resour. Res. 41, W07001, doi:10.1029/2004WR003333. Bertoldi, W., Tubino, M., 2006. River bifurcations: experimental observations on equilibrium conﬁgurations. Water Resour. Res., in press. Blondeaux, P., Seminara, G., 1985. A uniﬁed bar-bend theory of river meanders. J. Fluid Mech. 112, 363–377. Bolla Pittaluga, M., Federici, B., Repetto, R., et al., 2001. The morphodynamics of braiding rivers: experimental and theoretical results on unit processes. In: Mosley, M.P. (Ed.), Gravel-Bed Rivers V. New Zealand Hydrological Society, Wellington, pp. 143–181. Bolla Pittaluga, M., Repetto, R., Tubino, M., 2003. Channel bifurcation in braided rivers: Equilibrium conﬁgurations and stability. Water Resources Research 39(3), 1046, doi:10.1029/2001WR001112. Bristow, C., Best, J., 1993. Braided rivers: Perspectives and problems. In: Best, J.L. and Bristow, C.S. (Eds), Braided Rivers: Form, Process and Economic Applications. Geological Society Special Publication 75, 1–9. Bulle, H., 1926. Untersuchungen u¨ber die geschiebeableitung bei der spaltung von wasserla¨ufen. Technical Report, VDI Verlag, Berlin (in German). Chew, L., Ashmore, P., 2001. Channel adjustment and a test of rational regime theory in a proglacial braided stream. Geomorphology 37, 43–63. Colombini, M., Seminara, G., Tubino, M., 1987. Finite-amplitude alternate bars. J. Fluid Mech. 181, 213–232. de Heer, A., Mosselman, E., 2004. Flow structure and bedload distribution at alluvial diversions. In: Proceedings of RiverFlow2004, Napoli, 23–25 June. Exner, F., 1925. Uber die wechselwirkung zwischen wasser und geschiebe in ﬂussen. Sitzungsber. Akad. Wiss., Wein 165 (3–4), 165–203, (in German). Federici, P., Paola, C., 2003. Dynamics of bifurcations in noncohesive sediments. Water Resour. Res. 39 (6), 1162, doi:10.1029/2002WR001434. Ferguson, R., Ashmore, P., Ashworth, P., et al., 1992. Measurements in a braided river chute and lobe. Part 1. Flow pattern, sediment transport and channel change. Water Resour. Res. 28 (7), 1877–1886. Fredsoe, J., 1978. Meandering and braiding of rivers. J. Fluid Mech. 84, 607–624. Friedkin, J., 1945. A laboratory study of the meandering of alluvial rivers. Waterways Experimental Station Report. U.S. Army Corps of Engineers, Vicksburg, MS. Gilvear, D., 1999. Fluvial geomorphology and river engineering: future roles utilizing a ﬂuvial hydrosystem framework. Geomorphology 31, 229–245. Gran, K., Paola, C., 2001. Riparian vegetation controls on braided stream dynamics. Water Resour. Res. 37 (12), 3275–3284. Grifﬁths, G., 1981. Stable-channel design in gravel-bed rivers. J. Hydrol. 52, 291–305. Gurnell, A., Petts, G., Hannah, D., et al., 2001. Riparian vegetation and island formation along the gravelbed ﬁume tagliamento, Italy. Earth Surf. Process. Landf. 26, 31–62. Hirose, K., Hasegawa, K., Meguro, H., 2003. Experiments and analysis on mainstream alternation in a bifurcated channel in mountain rivers. In: Proceedings 3rd International Conference on River, Costal and Estuarine Morphodynamics, 1–5 September, Barcelona, Spain, pp. 571–583.

158

M. Tubino, W. Bertoldi

Hoey, T., 1992. Temporal variations in bedload transport rates and sediment storage in gravel-bed rivers. Prog. Phys. Geogr. 16 (3), 319–338. Hoey, T., Sutherland, A., 1991. Channel morphology and bedload pulses in braided rivers: a laboratory study. Earth Surf. Process. Landf. 16, 447–462. Ikeda, S., Parker, G., Sawai, K., 1981. Bend theory of river meanders. Part 1: Linear development. J. Fluid Mech. 112, 363–377. Jagers, H., 2003. Modelling planform changes of braided rivers. Ph.D. Thesis, University of Twente, The Netherlands. Kinoshita, R., Miwa, H., 1974. River channel formation which prevents downstream translation of transverse bar. Shinsabo (in Japanese) 94, 12–17. Klaassen, G., Douben, K., van der Waal, M., 2002 (4–6 September). Novel approaches in river engineering. In: Proceedings of RiverFlow 2002. Lovain la Neuve, Belgium, pp. 27–43. Klaassen, G., Masselink, G., 1992. Planform changes of a braided river with ﬁne sand as bed and bank material. In: Proceedings of 5th International Symposium on River Sedimentation, Vol. I, Karlsruhe, Germany, pp: 459–471. Krigstrom, A., 1962. Geomorphological studies of sandar plains and their braided rivers in iceland. Geogr. Ann. 44, 328–365. Kuroki, M., Kishi, T., 1985. Regime criteria on bars and braids. Report, Hokkaido University, Japan, Vol. 14, pp. 283–300. Lanzoni, S., Tubino, M., 1999. Grain sorting and bar instability. J. Fluid Mech. 393, 149–174. Leopold, L.B., Wolman, G., 1957. River channel patterns: braided, meandering and straight. United States Geological Survey Professional Paper 282B, pp. 39–85. Lisle, T., Ikeda, H., Iseya, F., 1991. Formation of stationary alternate bars in a steep channel with mixed size sediment: a ﬂume experiment. Earth Surf. Process. Landf. 16, 463–469. Miori, S., Repetto, R., Tubino, M., 2006. A one-dimensional model of bifurcations in gravel bed channels with erodible banks. Water Resour. Res. 42 (11), W11413. Mosley, M., 1983. Response of braided rivers to changing discharge. J. Hydrol. NZ 22, 18–67. Murray, B., Paola, C., 1994. A cellular model of braided rivers. Nature 371, 54–57. Paola, C., 2001. Modelling stream braiding over a range of scales. In: Mosley, M.P. (Ed.), Gravel-Bed Rivers V. New Zealand Hydrologic Society, Wellington, pp. 11–46. Parker, G., 1990. Surface-based bedload transport relation for gravel rivers. J. Hydraul. Res. 28, 417–436. Repetto, R., Tubino, M., 1999. Transition from migrating alternate bars to steady central bars in channels with variable width. In: Proceedings of International Symposium on River, Coastal and Estuarine Morphodynamics, Genova, Italy, 6–10 September. Repetto, R., Tubino, M., Paola, C., 2002. Planimetric instability of channels with variable width. J. Fluid Mech. 457, 79–109. Richardson, W., Thorne, C., 2001. Multiple thread ﬂow and channel bifurcation in a braided river: Brahmaputra-Jamuna River, Bangladesh. Geomorphology 38, 185–196. Seminara, G., Tubino, M., 1989. Alternate bars and meandering: Free, forced and mixed interactions. In: Ikeda, S. and Parker, G. (Eds), River Meandering. Water Resources Monographies, 12, pp. 267–320. Seminara, G., Tubino, M., 1992. Weakly nonlinear theory of regular meanders. J. Fluid Mech. 244, 257–288. Slingerland, R., Smith, N., 2004. River avulsions and their deposits. Ann. Rev. Earth Planet Sci. 32, 257–285. Southard, J., Smith, N., Kuhnle, R., 1984. Chutes and lobes: Newly identiﬁed elements of braiding in shallow gravelly streams. In: Koster, E.H. and Steel, R.J. (Eds), Sedimentology of Gravels and Conglomerates. Can. Soc. Petr. Geol. Mem. 10, pp. 51–59. Stojic, M., Chandler, J., Ashmore, P., Luce, J., 1998. The assessment of sediment transport rates by automated digital photogrammetry. Photogramm. Eng. Rem. Sens. 64 (5), 387–395. Tubino, M., Seminara, G., 1990. Free-forced interactions in developing meanders and suppression of free bars. J. Fluid Mech. 214, 131–159. Van der Nat, D., Schmidt, A., Tockner, K., Edwards, P., Ward, J., 2002. Inundation dynamics in braided ﬂoodplains: Tagliamento river, northeast Italy. Ecosystem 5, 636–647.

Bifurcations in gravel-bed streams

159

Wang, B., Fokking, R., De Vries, M., Langerak, A., 1995. Stability of river bifurcations in 1D morphodynamics models. J. Hydraul. Res. 33 (6), 739–750. Warburton, J., Davies, T., 1994. Variability of bedload transport and channel morphology in a braided river hydraulic model. Earth Surf. Process. Landf. 19, 403–421. Zolezzi, G., Bertoldi, W., Tubino, M., 2006. Morphological analysis and prediction of channel bifurcations. In: Sambrook-Smith, G.H., Best, J.L., Bristow, C.S., and Petts, G.E. (Eds), Braided Rivers: Process, Deposits, Ecology and Management. IAS Special Publication 36, Blackwell, Oxford, UK, pp. 233–256. Zolezzi, G., Guala, M., Termini, D., Seminara, G., 2005. Experimental observation of upstream overdeepening. J. Fluid Mech. 531, 191–219. Zolezzi, G., Seminara, G., 2001. Downstream and upstream inﬂuence in river meandering. Part 1. General theory and application of overdeepening. J. Fluid Mech. 438, 183–211.

Discussion by Rob Ferguson You have shown very elegantly that it is possible to consider the stability of bifurcations by a simple analytical approach, and to include the effects of differences in bed level between the two channels. Your assumption that all the ﬂow is within the channels seems very reasonable for avulsions and for mature bifurcations in braided rivers, but what about incipient bifurcation around a newly formed mid-channel bar in a braided river? In this case some of the ﬂow will be over the bar. Could this alter the stability of the bifurcation, and could it be included in your analytical approach?

Reply by the authors The analytical model presented here is obviously unable to describe the effect of over-bank ﬂow because it is essentially a 1D model based on the assumption that all the ﬂow is within the channels. Hence, the model may not be suitable to describe the dynamics of a bifurcation forming through the deposition of a mid-channel bar, as well as the analogous inception process discussed in Section 2 due to the interaction between bars and bank erosion. However, in gravel-bed braided rivers the ﬂow is typically shallow and sediments transport mainly occurs where the ﬂow ﬁeld is relatively intense, even under formative conditions. Hence, in active bifurcations morphological processes are mainly restricted to the deeper areas inside the branches and over-bank ﬂow is likely to play a minor role, at least at intermediate high stages. This may be also due to bed material sorting effects that typically produce an armoured layer. Recent observations performed on Tagliamento river support the above conjecture in that mayor changes in bed topography have been measured where a channelled ﬂow was clearly identiﬁable. Therefore, we do not expect the overall stability of a bifurcation to be strongly affected by over-bank ﬂow. In this respect we may note that our theoretical model is also able to reproduce, at least qualitatively, experimental ﬁndings with

Bifurcations in gravel-bed streams

159

Wang, B., Fokking, R., De Vries, M., Langerak, A., 1995. Stability of river bifurcations in 1D morphodynamics models. J. Hydraul. Res. 33 (6), 739–750. Warburton, J., Davies, T., 1994. Variability of bedload transport and channel morphology in a braided river hydraulic model. Earth Surf. Process. Landf. 19, 403–421. Zolezzi, G., Bertoldi, W., Tubino, M., 2006. Morphological analysis and prediction of channel bifurcations. In: Sambrook-Smith, G.H., Best, J.L., Bristow, C.S., and Petts, G.E. (Eds), Braided Rivers: Process, Deposits, Ecology and Management. IAS Special Publication 36, Blackwell, Oxford, UK, pp. 233–256. Zolezzi, G., Guala, M., Termini, D., Seminara, G., 2005. Experimental observation of upstream overdeepening. J. Fluid Mech. 531, 191–219. Zolezzi, G., Seminara, G., 2001. Downstream and upstream inﬂuence in river meandering. Part 1. General theory and application of overdeepening. J. Fluid Mech. 438, 183–211.

Discussion by Rob Ferguson You have shown very elegantly that it is possible to consider the stability of bifurcations by a simple analytical approach, and to include the effects of differences in bed level between the two channels. Your assumption that all the ﬂow is within the channels seems very reasonable for avulsions and for mature bifurcations in braided rivers, but what about incipient bifurcation around a newly formed mid-channel bar in a braided river? In this case some of the ﬂow will be over the bar. Could this alter the stability of the bifurcation, and could it be included in your analytical approach?

Reply by the authors The analytical model presented here is obviously unable to describe the effect of over-bank ﬂow because it is essentially a 1D model based on the assumption that all the ﬂow is within the channels. Hence, the model may not be suitable to describe the dynamics of a bifurcation forming through the deposition of a mid-channel bar, as well as the analogous inception process discussed in Section 2 due to the interaction between bars and bank erosion. However, in gravel-bed braided rivers the ﬂow is typically shallow and sediments transport mainly occurs where the ﬂow ﬁeld is relatively intense, even under formative conditions. Hence, in active bifurcations morphological processes are mainly restricted to the deeper areas inside the branches and over-bank ﬂow is likely to play a minor role, at least at intermediate high stages. This may be also due to bed material sorting effects that typically produce an armoured layer. Recent observations performed on Tagliamento river support the above conjecture in that mayor changes in bed topography have been measured where a channelled ﬂow was clearly identiﬁable. Therefore, we do not expect the overall stability of a bifurcation to be strongly affected by over-bank ﬂow. In this respect we may note that our theoretical model is also able to reproduce, at least qualitatively, experimental ﬁndings with

Bifurcations in gravel-bed streams

159

Wang, B., Fokking, R., De Vries, M., Langerak, A., 1995. Stability of river bifurcations in 1D morphodynamics models. J. Hydraul. Res. 33 (6), 739–750. Warburton, J., Davies, T., 1994. Variability of bedload transport and channel morphology in a braided river hydraulic model. Earth Surf. Process. Landf. 19, 403–421. Zolezzi, G., Bertoldi, W., Tubino, M., 2006. Morphological analysis and prediction of channel bifurcations. In: Sambrook-Smith, G.H., Best, J.L., Bristow, C.S., and Petts, G.E. (Eds), Braided Rivers: Process, Deposits, Ecology and Management. IAS Special Publication 36, Blackwell, Oxford, UK, pp. 233–256. Zolezzi, G., Guala, M., Termini, D., Seminara, G., 2005. Experimental observation of upstream overdeepening. J. Fluid Mech. 531, 191–219. Zolezzi, G., Seminara, G., 2001. Downstream and upstream inﬂuence in river meandering. Part 1. General theory and application of overdeepening. J. Fluid Mech. 438, 183–211.

Discussion by Rob Ferguson You have shown very elegantly that it is possible to consider the stability of bifurcations by a simple analytical approach, and to include the effects of differences in bed level between the two channels. Your assumption that all the ﬂow is within the channels seems very reasonable for avulsions and for mature bifurcations in braided rivers, but what about incipient bifurcation around a newly formed mid-channel bar in a braided river? In this case some of the ﬂow will be over the bar. Could this alter the stability of the bifurcation, and could it be included in your analytical approach?

Reply by the authors The analytical model presented here is obviously unable to describe the effect of over-bank ﬂow because it is essentially a 1D model based on the assumption that all the ﬂow is within the channels. Hence, the model may not be suitable to describe the dynamics of a bifurcation forming through the deposition of a mid-channel bar, as well as the analogous inception process discussed in Section 2 due to the interaction between bars and bank erosion. However, in gravel-bed braided rivers the ﬂow is typically shallow and sediments transport mainly occurs where the ﬂow ﬁeld is relatively intense, even under formative conditions. Hence, in active bifurcations morphological processes are mainly restricted to the deeper areas inside the branches and over-bank ﬂow is likely to play a minor role, at least at intermediate high stages. This may be also due to bed material sorting effects that typically produce an armoured layer. Recent observations performed on Tagliamento river support the above conjecture in that mayor changes in bed topography have been measured where a channelled ﬂow was clearly identiﬁable. Therefore, we do not expect the overall stability of a bifurcation to be strongly affected by over-bank ﬂow. In this respect we may note that our theoretical model is also able to reproduce, at least qualitatively, experimental ﬁndings with

160

M. Tubino, W. Bertoldi

non-channelled bifurcating ﬂows. In particular, theoretical results highlight the role of Shields stress of incoming ﬂow as the crucial parameter affecting ﬂow partition at the bifurcation, in agreement with Federici and Paola (2003) observations. Furthermore, the stable symmetrical conﬁgurations that have been invariably observed by Ashworth (1996) are in fairly good agreement with the sub-critical character of his experimental runs.

Gravel-Bed Rivers VI: From Process Understanding to River Restoration H. Habersack, H. Pie´gay, M. Rinaldi, Editors r 2008 Elsevier B.V. All rights reserved.

161

7 The importance of ﬂoods for bed topography and bed sediment composition: numerical modelling of Rhine bifurcation at Pannerden Erik Mosselman and Kees Sloff

Abstract The morphological response of the Dutch Rhine branches to interventions depends sensitively on the morphological development of the bifurcations. Computations in the 1990s revealed that proper modelling of this morphological development requires the inclusion of physical mechanisms for grain sorting. Therefore we apply a twodimensional (2D) morphological Delft3D model with graded sediment to the Rhine bifurcation at Pannerden (‘‘Pannerdensche Kop’’). We ﬁnd that such a computation, unlike computations with uniform sediment, hardly produces any changes in bed topography at a discharge of 2400 m3/s. Only a less frequent ﬂood discharge of 6000 m3/s is found to produce topography changes. Apparently, the highest ﬂoods need to be included when modelling the combined evolution of bed topography and bed sediment composition. We explain this from the time scales for bed topography development and the development of bed sediment composition. If the composition pattern develops fast with respect to bed topography, the sediment is rearranged immediately in a way that eliminates the gradients in sediment transport capacity, after which the bed topography remains unchanged. We show theoretically that the ratio of transport layer thickness to ﬂow depth is the key parameter that determines the ratio of these time scales. The ﬁndings also suggest that existing theories tend to underestimate the thickness of the transport layer, with the important implication that computational results may give the false idea of a stable bed topography in cases where the real topography is not stable at all. 1.

Introduction

The river Rhine divides into three branches in the Netherlands (Fig. 7.1). The morphological stability of these branches depends on how sediment transport is distributed over the branches at the Pannerdensche Kop and IJsselkop bifurcations. E-mail addresses: [email protected] (E. Mosselman), [email protected] (K. Sloff) ISSN: 0928-2025

DOI: 10.1016/S0928-2025(07)11124-X

162

E. Mosselman, K. Sloff

Figure 7.1. Location of Pannerdensche Kop bifurcation of river Rhine, the Netherlands.

If the distribution of sediment transport is perturbed, branches short of sediment will erode and branches overfed with sediment will experience sedimentation. This eventually changes the discharges of the branches, thus having important implications for ﬂooding risks, navigation depths and regional water supply. Wang et al. (1995) demonstrate that the dependence of downstream branches on the distribution of sediment transport at bifurcations can be very sensitive. Currently the two Rhine bifurcations are more or less stable, despite a small trend of diminishing discharges in the Waal branch. However, extensive interventions are being prepared to comply with a new policy for safety against ﬂooding. This policy seeks to provide more space to the river rather than to continually raise the embankments in response to climate change scenarios and improved risk assessments. Interventions to create more room include the removal of bottlenecks, the lowering of ﬂoodplains, the modiﬁcation of groynes and the excavation of secondary channels. They will affect sediment transport and morphology, including the distribution of sediment transport at the bifurcations. The impacts can be assessed using a numerical morphological model. In the 1980s, application of a two-dimensional (2D) morphological model to the Pannerdensche Kop bifurcation was seen as a great success, because the bed topography computed by the model agreed better with prototype measurements than the bed topography obtained in a mobile-bed physical model (Fig. 7.2). Those computations had been carried out by assuming, for each of the branches separately, spatially constant sediment properties and a spatially constant Che´zy coefﬁcient for hydraulic resistance. Samples of river bed material indicated, however, that sediment granulometry varies spatially due to grain sorting. The computations were therefore

The importance of floods for bed topography and bed sediment composition

163

Figure 7.2. Morphological modelling of Pannerdensche Kop bifurcation in the 1980s, with bed topography from ﬁeld measurements (top), scale model (centre) and 2D numerical model (bottom).

repeated at the end of the 1990s using spatially varying grain sizes, albeit without physics-based process descriptions for changes in bed sediment composition. Surprisingly, it appeared no longer possible to reproduce the bed topography correctly. The topography was found to be affected signiﬁcantly by both spatial grain size variations and spatial variations in hydraulic roughness (Mosselman et al., 1999,

E. Mosselman, K. Sloff

164

2003). The success of computations using spatially constant sediment properties and a spatially constant Che´zy coefﬁcient was ascribed to the fact that the effect of spatial grain size variations is more or less counterbalanced by the effect of spatial roughness variations. However, there are no reasons to assume that this counterbalancing is exact or that it occurs under other conditions as well. Proper modelling was therefore considered possible only if physical mechanisms for grain sorting and alluvial roughness would be included in the model. Given the importance of grain sorting, we apply a 2D morphological model with formulations of physical mechanisms for the transport of graded sediment to the Pannerdensche Kop bifurcation, where the main branch of the Rhine, called ‘‘BovenRijn’’, divides into the Waal to the left and the Pannerdensch Kanaal to the right (Fig. 7.1). The riverbed in this area consists of coarse sands and ﬁne gravels. We investigate the performance of the model at average and ﬂood discharges. Remarkably, the bed topography hardly changes under the average discharge despite substantial sediment transport in the computation. Only a high, less frequent ﬂood discharge is found to produce changes in bed topography, thus questioning the common opinion that a representative annual discharge hydrograph or even a single bed-forming discharge close to bankfull provides sufﬁcient information on the hydrodynamic conditions that cause long-term morphological change. We explain this from the time scales for bed topography development and development of bed sediment composition. If the sediment composition pattern develops fast with respect to the development of bed topography, it is the sediment composition that responds immediately to the initial conditions in a way that eliminates the gradients in sediment transport capacity, after which the bed topography remains unchanged. We show theoretically that this occurs when the thickness of the active or transport layer on the bed is much smaller than the ﬂow depth. A combined evolution of bed topography and bed sediment composition is only possible if the two time scales have the same order of magnitude, i.e., if the active-layer thickness has the same order of magnitude as about one ﬁfth of the ﬂow depth. As this occurs during the larger, less frequent ﬂoods, it demonstrates the importance of ﬂoods for bed topography and bed sediment composition in the Dutch Rhine branches at Pannerden. We argue, however, that the explanation based on ﬂoods may not be sufﬁcient. It is also possible that textbook values for active-layer thickness underestimate the true values that should be adopted in the model.

2. 2.1.

Numerical model Two-dimensional river morphodynamics

The Delft3D modules for 2D river morphodynamics are based on depth-averaged, steady-ﬂow equations, a volumetric sediment balance and formulae relating magnitude and direction of sediment transport to local ﬂow ﬁeld and bed topography. The ﬁrst 2D model of this kind was developed by Van Bendegom (1947), but its application without the modern computer devices of the present was very laborious.

The importance of floods for bed topography and bed sediment composition

165

Subsequent work on helical ﬂow in river bends (e.g., Kalkwijk and De Vriend, 1980) and forces on sediment grains on a transversely sloping bed (e.g., Engelund, 1974; Odgaard, 1981) resulted in the model of Struiksma (1985), Struiksma et al. (1985) and Olesen (1987). Similar models were developed by Shimizu and Itakura (1985) and Nelson and Smith (1989). The earlier ﬂow and bed topography model by Kennedy et al. (1984) does not fall into this class of models since it does not use the sediment balance. Instead, it uses an axisymmetrical relationship to link the bed topography directly to the local ﬂow ﬁeld. Struiksma et al. (1985) also developed important insights for proper calibration and veriﬁcation of numerical morphological models. They found that the bed topography in bends of ﬂumes and rivers can be understood as a superimposition of a uniform transversely sloping bed and a pattern of non-migrating alternate bars that attenuates in the tail of the bend. Reproduction of the attenuated alternate bars in the tail of the bend is the main indicator that the model can properly reproduce bed topographies in rivers with arbitrary geometries. However, a model that fails to reproduce these attenuated bars can still produce individual cross-sections that agree well with observations. A major pitfall is hence that 2- and 3-D morphological models are calibrated and veriﬁed against individual cross-sections only, without considering the spatial patterns of bars and pools. Historically, the Delft3D modules for 2D river morphodynamics can be seen as direct descendants of the models by Van Bendegom (1947), Struiksma (1985) and Struiksma et al. (1985), but with additional features, such as the transport of graded sediment. Moreover, the integration of these models into a larger software system has led to several modiﬁcations in the hydrodynamic modules. To assure that such modiﬁcations do not deteriorate the performance of the model, each new version of Delft3D is validated on the basis of a test bank containing 70 cases of laboratory experiments and exact analytical solutions, including speciﬁc cases of river morphodynamics. Lesser et al. (2004) present the full hydrodynamic equations of Delft3D. These equations are quasi-3D in the sense that the vertical momentum equation has been reduced to the hydrostatic pressure equation by assuming that vertical ﬂow accelerations are negligible compared to gravity. The 3D model thus consists of several 2D layers that are coupled through the hydrostatic pressure equation and a continuity equation for mass conservation. This allows an approach in which the horizontal sizes of the computational grid are much larger than the vertical sizes. This suits the modelling of natural water bodies such as rivers, because in these systems the horizontal extent of the computational domain is usually much larger than the water depth. We use a single layer in our morphological computations, which means that we use 2D hydrodynamic equations despite the availability of quasi-3D equations in the software. We do this because morphological computations require a new computation of the ﬂow ﬁeld after each time step of bed evolution, leading to long computation times. Fast computation of the ﬂow ﬁeld greatly improves the overall performance of the model. The essential 3D feature of helical ﬂow is included in the 2D equations by means of a parameterization. Helical ﬂow arises in curved ﬂows and produces a difference between the near-bed ﬂow direction and the depth-averaged

E. Mosselman, K. Sloff

166

ﬂow direction. In rivers with predominantly bedload, the near-bed ﬂow direction is equal to the direction of sediment transport over a ﬂat bed. The equilibrium intensity of the helical ﬂow, I e , is calculated by Ie ¼

hu R

(7.1)

where h denotes ﬂow depth, u is the depth-averaged ﬂow velocity and R is the radius of streamline curvature deﬁned by 1 1 @ur ¼ R us @s

(7.2)

where us is the ﬂow velocity component along the streamline, ur the ﬂow velocity component perpendicular to the streamline and s a local coordinate along the streamline. An advection–diffusion equation describes how the actual helical ﬂow intensity, I, adapts to the equilibrium helical ﬂow intensity, Ie. The angle, at , between the near-bed ﬂow direction and the depth-averaged ﬂow direction is calculated from the actual helical ﬂow direction by pﬃﬃﬃ g I 2 tan at ¼ 2 1 (7.3) 2kC u k where k is the Von Karman constant (k ¼ 0:4), g is the acceleration due to gravity (g ¼ 9.8 m/s2) and C is the Che´zy coefﬁcient for hydraulic roughness. Sediment transport is described by semi-empirical transport formulae and a depthaveraged sediment balance. The latter reads ð1 Þ

@zb @qsx @qsy þ þ ¼0 @t @x @y

(7.4)

in which qsx and qsy are volumetric sediment transport components per unit width (excluding pores) in the x- and y-direction, respectively, t is time, x and y are horizontal space co-ordinates, zb denotes bed level and e is the porosity of the bed. The porosity factor translates the net volume of sediment grains into the bulk volume which corresponds to the volumetric changes of bed topography due to erosion and sedimentation. Usually e ¼ 0.4 is assumed (e.g., Jansen et al., 1979). Similar balance equations can be written for individual size fractions of a mixture of graded sediment. This is elaborated in the next section. The transport formula for graded sediment is presented in the next section as well. Besides helical ﬂow, transverse bed slopes also cause a difference between the directions of bedload and depth-averaged ﬂow. This effect is modelled by tan as ¼

sin at ð1=f Þð@zb =@yÞ cos at ð1=f Þð@zb =@xÞ

(7.5)

where as is the angle between the sediment transport direction and the depthaveraged ﬂow direction, and f is a dimensionless parameter. The next section presents the relation for f in case of graded sediment.

The importance of floods for bed topography and bed sediment composition 2.2.

167

Graded sediment

Graded sediments are accounted for through (i) division of the sediment mixture into separate fractions, (ii) transport formulae and mass conservation equations for each of the separate fractions, (iii) hiding-and-exposure corrections for the critical shear stress of each of the fractions, (iv) an active layer or transport layer affected by erosion and sedimentation (Hirano, 1972) and (v) a bookkeeping system for substratum that has become inactive due to sedimentation. The relative occurrence of a sediment size fraction i in the active layer is indicated by pi;a , and in the substratum by pi;0 . By deﬁnition X X pi;a ¼ 1 and pi;0 ¼ 1 (7.6) i

i

The mass conservation equation or sediment balance for each fraction reads ( pi;a sedimentation @pi;a d @z0 @qsxi @qsyi þ pi ðz0 Þ þ þ ¼ 0 pi ðz0 Þ ¼ ð1 Þ pi;0 erosion @t @t @x @y (7.7) in which z0 is the upper level of the substratum, pi ðz0 Þ is the relative occurrence of a sediment size fraction i at this level (taken equal to pi;a during sedimentation and equal to pi;0 during erosion), d is the thickness of the active layer, and qsxi and qsyi are volumetric bedload transport components per unit width for fraction i in the x- and y-direction, respectively. The actual bed level is given by zb ¼ z 0 þ d

(7.8)

The sediment transport rate per fraction is described using a standard transport formula. Here the formula of Meyer-Peter and Mu¨ller (1948) is taken as a starting point. For graded sediment, this formula is written as !3=2 3=2 qﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ C t b 0:047xi qsi ¼ Acal gDD3i rw gDDi C 90

(7.9)

where qsi is the total bedload transport rate per unit width for fraction i, Acal is a calibration factor (equal to 8 in Meyer-Peter and Mu¨ller’s original formula), D is the relative density of the sediment under water, deﬁned by D ¼ (rsrw)/rw, rw and rs are mass densities of water and sediment, respectively, Di is the grain size of fraction i, C90 is a Che´zy coefﬁcient for grain roughness, tb is the magnitude of the bed shear stress and xi is the hiding-and-exposure correction. The grain roughness is based on a Nikuradse roughness of 3D90 instead of Meyer-Peter and Mu¨ller’s original D90: 12h (7.10) C 90 ¼ 18 log 3D90 where D90 denotes the grain size exceeded by 10090 ¼ 10% of the sediment mixture.

E. Mosselman, K. Sloff

168

The hiding-and-exposure correction is modelled according to the Egiazaroff (1965) formulation, adjusted by Ashida and Michiue (1972, 1973) Dm xi ¼ 0:85 Di 2 32 log 19 5 xi ¼ 4 log 19 DDmi

for

Di o0:47 Dm

for

Di 0:47 Dm

(7.11)

where Dm denotes the average grain size of the sediment mixture. Equation (7.5) for the effect of transverse bed slopes is applied to calculate the direction of sediment transport for each size fraction separately. The corresponding parameters, fi, are calculated by Bsh C sh Dsh tb Di Dm fi ¼ Ash (7.12) rw gDDi h Di in which Ash , Bsh , C sh and Dsh are calibration parameters. However, as very little is known about the proper formulation for graded sediments, the distinction between different values for different size fractions is switched off by setting Csh ¼ 0 and Dsh ¼ 0. The usual value for Bsh is 0.5. The implementation of the graded-sediment equations in Delft3D has been validated against Ribberink’s (1987) straight-ﬂume experiments and Olesen’s (1987) curved-ﬂume experiments (Sloff et al., 2001). 2.3.

Model settings and initial and boundary conditions

We applied a spatially uniform Che´zy value of 45 m1/2/s. The sediment mixture was divided into the six size fractions presented in Table 7.1. The representative grain sizes, Di, for each fraction were calculated as the geometric mean of the upper and lower size limits. We used the transport formula of Meyer-Peter and Mu¨ller with a calibration coefﬁcient, Acal, equal to 11.2. The porosity of the bed was taken as equal to 0.4. The effect of transverse bed slopes on the direction of sediment transport was represented by Ash ¼ 0.8, Bsh ¼ 0.5, Csh ¼ 0 and Dsh ¼ 0. Table 7.1.

Division of sediment mixture into six separate size fractions.

Fraction number

1 2 3 4 5 6

Grain sizes (mm) Lower limit

Upper limit

Geometric mean (Di)

0.06 1 2 2.8 4 8

1 2 2.8 4 8 12.5

0.24 1.41 2.37 3.35 5.66 10.00

The importance of floods for bed topography and bed sediment composition

169

Two cases were computed, one with a constant discharge of 2400 m3/s and one with a constant discharge of 6000 m3/s. The corresponding Boven-Rijn water depths were on the order of 6 m and 11 m respectively. The 2400 m3/s discharge represents more or less the annual bed-forming conditions for the sand-bed Rhine branches further downstream. The 6000 m3/s discharge represents ﬂood conditions that actually occur at a Rhine discharge of 6700 m3/s because the residual 700 m3/s pass over the ﬂoodplains that were not included in the model. Water levels at the downstream boundaries were calculated with rating curves derived from a 1D SOBEK model of the Rhine branches. The initial bed topography was composed as an average of the topographies measured annually in the years 1988–1991, because none of the individual measured topographies covered the full model area (top of Fig. 7.3). The thickness of the active layers was estimated from dune height measurements analysed by Wilbers and Ten Brinke (2003). This resulted in a 0.1 m thickness at 2400 m3/s and a 1.0 m thickness at 6000 m3/s. The bed sediment composition was derived from measurements by Gruijters et al. (2001). An initial bed sediment composition for the computations was obtained by averaging the mean grain sizes over width and by assigning uniform distributions of the resulting grain sizes to each cross-section. This schematized initial bed sediment composition showed long-stream variations with a marked segregation between sediments in the Waal and the Pannerdensch Kanaal but no 2D patterns within the branches (bottom of Fig. 7.3). Each computation simulated the development during a period of 100 days.

3.

Results and analysis

Figure 7.4 shows the results for a 2400 m3/s discharge, Fig. 7.5 those for a 6000 m3/s discharge. Remarkably, the bed topography hardly changed under the 2400 m3/s discharge despite substantial sediment transport in the computation. The fast development of a patchy sediment composition pattern provides an explanation for the lack of bed level changes, because it immediately reduces all gradients in sediment transport capacity to zero. The slower development of sediment composition under the 6000 m3/s discharge does not eliminate the gradients in transport capacity. The result is a combined evolution of both bed topography and bed sediment composition that produces a much smoother sediment composition pattern. This smoother composition pattern complies better with the measured sediment composition in Fig. 7.6. These results reveal the importance of the ratio of the time scales for bed topography development and development of bed sediment composition. This ratio can be derived from the following theoretical analysis along the lines of Ribberink (1987). The 1D quasi-steady ﬂow equations read u

@u @zb @h gujuj þg þ 2 ¼0 þg @x @x @x C h

(7.13)

h

@u @h þu ¼0 @x @x

(7.14)

E. Mosselman, K. Sloff

170

6

4000

5 distance (m) →

3500 4 3000 3 2500

2

1

2000 2000

2500

3000 3500 distance (m) →

4000

4500

0

0.005

4000

0.0045 distance (m) →

3500 0.004 3000 0.0035 2500

0.003

0.0025

2000 2000

2500

3000 3500 distance (m) →

4000

4500

0.002

Figure 7.3. Initial bed topography (top, in m above NAP) and mean sediment grain sizes (bottom, in m) at Pannerdensche Kop.

The importance of floods for bed topography and bed sediment composition

171

6

4000

5 distance (m) →

3500 4 3000 3 2500

2

1

2000 2000

2500

3000 3500 distance (m) →

4000

4500

0

0.005

4000

0.0045 distance (m) →

3500 0.004 3000 0.0035 2500

0.003

0.0025

2000 2000

2500

3000 3500 distance (m) →

4000

4500

0.002

Figure 7.4. Bed topography (top, in m above NAP) and mean sediment grain sizes (bottom, in m) at Pannerdensche Kop resulting from computation with 2400 m3/s.

E. Mosselman, K. Sloff

172

6

4000

5 distance (m) →

3500 4 3000 3 2500

2

1

2000 2000

2500

3000 3500 distance (m) →

4000

4500

0

0.005

4000

0.0045 distance (m) →

3500 0.004 3000 0.0035 2500

0.003

0.0025

2000 2000

2500

3000 3500 distance (m) →

4000

4500

0.002

Figure 7.5. Bed topography (top, in m above NAP) and mean sediment grain sizes (bottom, in m) at Pannerdensche Kop resulting from computation with 6000 m3/s.

The importance of floods for bed topography and bed sediment composition

173

Figure 7.6. Measured sediment grain sizes. (After Gruijters et al. (2001)).

The friction term in equation (7.13) can be neglected when focusing the attention on short spatial scales. Substitution of equation (7.14) into equation (7.13) then leads to the simpliﬁed ﬂow equation @u u @zb ¼0 ¼ 2 @x hð1 Fr Þ @x

(7.15)

pﬃﬃﬃﬃﬃ in which the Froude number, Fr, is deﬁned by Fr ¼ u= gh. For a 1D system, the sediment balance for the complete sediment mixture in equation (7.4) can be written as @zb @qs þ ¼0 (7.16) @t @x in which qs is the volumetric sediment transport rate per unit width (excluding pores). The sediment transport is a function of both ﬂow velocity, u, and average sediment grain size, Dm: ð1 Þ

@qs dqs @u dqs @Dm ¼ þ @x du @x dDm @x

(7.17)

Substitution of this equation into equation (7.16) and elimination of @u=@x by applying equation (7.15) lead to @zb u dqs @zb 1 dqs @Dm þ ¼ @t ð1 Þð1 Fr2 Þh du @x ð1 Þ dDm @x

(7.18)

This equation can be interpreted as a kinematic bed topography wave forced by gradients in sediment composition. The corresponding celerity is cbed ¼

u dqs ð1 Þð1 Fr2 Þh du

(7.19)

This relation assumes a simpler form by introducing the degree of nonlinearity, b, of qs ¼ qs ðuÞ, which is deﬁned by b ¼ ðu=qs Þðdqs =duÞ. The result reads cbed ¼

bqs ð1 Þð1 Fr2 Þh

(7.20)

E. Mosselman, K. Sloff

174 Equation (7.7) reduces in a 1D system to @pi;a d @z0 @q þ pi ðz0 Þ þ si ¼ 0 ð1 Þ @t @t @x

(7.21)

The active layer thickness, d, is assumed constant, so that @z0 =@t ¼ @zb =@t. The transport rate per fraction is written as qsi ¼ piT qs in which piT is the relative occurrence of sediment size fraction i in the bedload. Subsequently equation (7.16) is used to eliminate @qs =@x. The result is @p @zb @p (7.22) ð1 Þ d i;a þ ðpi ðz0 Þ piT Þ þ qs iT ¼ 0 @t @t @x Multiplication of all terms by Di and summation over all size fractions gives, assuming that the substratum has the same composition as the active layer ðpi ðz0 Þ ¼ pi;a Þ and noting that the values of Di are constants for the selected size fractions: @Dm @zb @DmT þ ðDm DmT Þ þ qs ¼0 (7.23) ð1 Þ d @t @t @x with Dm ¼

X

pi;a Di

(7.24)

i

DmT ¼

X

piT Di

(7.25)

i

Assuming a constant ratio of average bedload grain size to average active-layer grain size, m ¼ DmT =Dm , equation (7.23) can be written as @Dm mqs @Dm DmT Dm @zb þ ¼ @t dð1 Þ @x d @t

(7.26)

This equation can be interpreted as a kinematic bed sediment composition wave forced by bed level changes and a difference between the sediment compositions of the bedload and the bed. The corresponding celerity is mqs (7.27) cmix ¼ dð1 Þ The celerities in equations (7.20) and (7.27) deﬁne the time scales of morphological change and change in sediment composition. The ratio of these time scales reads T mix cbed b d ¼ ¼ T bed cmix mð1 Fr2 Þ h

(7.28)

In sand bed rivers, b has values of 3 to 5, but higher values are possible in conditions close to the initiation of sediment motion (Paintal, 1971). Furthermore, m is an order 1 parameter and Fr2 1 in the Dutch Rhine branches. Hence T mix d 5 T bed h

(7.29)

The importance of floods for bed topography and bed sediment composition

175

The ratio of time scales is hence proportional to the ratio of active-layer thickness to ﬂow depth. If d h, the sediment composition of the bed responds immediately to the initial bed topography in a way that eliminates the gradients in sediment transport capacity, after which the bed topography remains unchanged. This is the case for the 2400 m3/s discharge with d ¼ 0.1 m. If d h, the evolution of the bed topography is forced by the initial sediment composition pattern, while the bed sediment composition remains unchanged. This resembles the nature of the computations in the 1990s using spatially varying grain sizes without physics-based process descriptions for changes in bed sediment composition. A combined evolution of bed topography and bed sediment composition is only possible if (ﬁve times) d and h have the same order of magnitude. This is the case for the 6000 m3/s discharge with d ¼ 1 m.

4.

Discussion and conclusion

The results and the analysis demonstrate the importance of the ratio of active-layer thickness to ﬂow depth. A combined evolution of bed topography and bed sediment composition is only possible if the active-layer thickness has the same order of magnitude as about one ﬁfth of the ﬂow depth. In the Dutch Rhine branches, such conditions occur during the large but rare ﬂoods that can be ignored in discharge hydrograph schematizations for traditional morphological computations with uniform sediment. Apparently these ﬂoods must be accounted for when simulating the combined evolution of topography and composition. At lower ﬂows, the bed topography produced by the previous ﬂood forces a rapid re-arrangement of grain sizes that suppresses further bed level changes. However, inclusion of high ﬂoods still may not yield the full picture. The observation from measurements that erosion and sedimentation in the Pannerdensche Kop area occur at lower discharges as well, casts doubt on the common wisdom that the active-layer thickness is equal to half the dune height, or somewhat larger due to sporadic deeper troughs. This common wisdom seems to underestimate the true thickness of the active layer to be adopted in the model. Better estimates might result from the new depth-continuous modelling approach by Blom (2003) and Blom et al. (2003). Moreover, the actual active layer on larger time scales depends on all kinds of local bed level ﬂuctuations generated by discharge variations, i.e., not on dunes alone. Examples are the sand waves generated by fast changes in discharge, the sand waves generated by the ﬂow pattern changes due to the ﬂooding of ﬂoodplains, and the stage-dependent variations of the transverse bed slope in river bends. Increased shear stresses due to navigation during low ﬂows might have an effect as well. The ﬁnal conclusion is that the thickness of the active layer is a key parameter in morphological modelling with graded sediment. Proper representation of its effect implies that ﬂoods need to be included when modelling the combined evolution of bed topography and bed sediment composition, even in the absence of armouring. At the same time, however, textbook values seem to underestimate the true thickness of active layers. An important implication is that the application of standard theories might easily lead to the false idea of a stable bed topography that, in reality, is not stable at all. This is an important topic for further research.

E. Mosselman, K. Sloff

176

Acknowledgements This study has been carried out with Doelﬁnanciering funds from the Dutch Ministry of Transport, Public Works and Water Management. It is part of a joint research programme of the Morphological Triangle, a thematic subdivision of the Netherlands Centre for River Studies (NCR) that co-ordinates the ﬂuvial morphological research in the Netherlands. The data have been provided by Rijkswaterstaat RIZA. The set-up of the computations has been based on numerous previous morphological computations by Michele Bernabe` of the University of Trento, Italy, and Patrick Verhaar of Delft University of Technology, the Netherlands.

Appendix 1: List of symbols Acal Ash Bsh b C Csh C90 cbed cmix Di Dm DmT Dsh D90 Fr f fi g h I Ie pi(z0) pi,a pi,0 piT

calibration factor in sediment transport formula () coefﬁcient in relation for effect of transverse bed slopes on sediment transport direction () exponent of Shields parameter in relation for effect of transverse bed slopes on sediment transport direction () degree of nonlinearity in qs ¼ qs ðuÞ (), deﬁned by b ¼ ðu=qs Þðdqs =duÞ Che´zy coefﬁcient for hydraulic roughness (m1/2/s) exponent of Di =h in relation for effect of transverse bed slopes on sediment transport direction () Che´zy coefﬁcient for grain roughness (m1/2/s) celerity of inﬁnitesimal perturbations in bed topography (m/s) celerity of inﬁnitesimal perturbations in bed sediment composition (m/s) sediment grain size of fraction i (m) average grain size of sediment mixture (m) average grain size of bedload (m) exponent of Dm =Di in relation for effect of transverse bed slopes on sediment transport direction () 90% percentile of sediment grain pﬃﬃﬃﬃﬃ size distribution (m) Froude number (), Fr ¼ u= gh parameter weighing the inﬂuence of gravity pull along transverse bed slopes on transport direction of complete sediment mixture () parameter weighing the inﬂuence of gravity pull along transverse bed slopes on transport direction of sediment size fraction i () acceleration due to gravity (m/s2) ﬂow depth (m) local instantaneous intensity of helical ﬂow (m/s) equilibrium intensity of helical ﬂow (m/s) relative occurrence of sediment size fraction i at upper level of substratum, z0 () relative occurrence of sediment size fraction i in active layer () relative occurrence of sediment size fraction i in substratum () relative occurrence of sediment size fraction i in bedload ()

The importance of floods for bed topography and bed sediment composition qs qsi qsx qsxi qsy qsyi R s Tbed Tmix t u ur us x y zb z0 as at D d e k m xi rs rw tb

177

volumetric sediment transport rate per unit width for complete sediment mixture (m2/s) volumetric sediment transport rate per unit width for sediment size fraction i (m2/s) volumetric sediment transport component per unit width in x-direction for complete sediment mixture (m2/s) volumetric sediment transport component per unit width in x-direction for sediment size fraction i (m2/s) volumetric sediment transport component per unit width in y-direction for complete sediment mixture (m2/s) volumetric sediment transport component per unit width in y-direction for sediment size fraction i (m2/s) radius of streamline curvature (m) local coordinate along streamline (m) timescale of changes in bed topography (s) timescale of changes in bed sediment composition (s) time (s) depth-averaged ﬂow velocity (m/s) depth-averaged ﬂow velocity component perpendicular to streamline (m/s) depth-averaged ﬂow velocity component along streamline (m/s) horizontal coordinate (m) horizontal coordinate (m) bed level (m+datum) upper level of substratum (m+datum) angle between sediment transport direction and depth-averaged ﬂow direction (rad) angle between near-bed ﬂow direction and depth-averaged ﬂow direction (rad) relative density of sediment under water (), D ¼ ðrs rw Þ=rw thickness of active layer (m) porosity of bed () Von Karman constant () ratio of average grain size of bedload to average grain size of active layer (), m ¼ DmT/Dm hiding-and-exposure correction () mass density of sediment (kg/m3) mass density of water (kg/m3) magnitude of bed shear stress (N/m2)

References Ashida, K., Michiue, M., 1972. Study on hydraulic resistance and bedload transport rate in alluvial streams. Trans. Japn. Soc. Civil Eng. 206, 59–69. Ashida, K., Michiue, M., 1973. Studies on bed-load transport in open channel ﬂows. Proceedings of International Symposium on River Mechanics, IAHR, Bangkok, Thailand, Paper No. A36, pp. 407–418.

178

E. Mosselman, K. Sloff

Blom, A., 2003. A continuum vertical sorting model for rivers with non-uniform sediment and dunes. Ph.D. Thesis, Twente University, Enschede, The Netherlands. Blom, A., Ribberink, J.S., Parker, G., 2003. Sediment continuity for rivers with non-uniform sediment, dunes, and bed load transport. In: Gyr, A. and Kinzelbach, W. (Eds), Sedimentation & Sediment Transport: At the Crossroad of Physics and Engineering. Kluwer Academic Publishers, Switzerland, pp. 179–182. Egiazaroff, I.V., 1965. Calculation of non-uniform sediment concentrations. J. Hydraul. Div. ASCE 91 (HY4), 225–247. Engelund, F., 1974. Flow and bed topography in channel bends. J. Hydraul. Div. ASCE 100 (HY11), 1631–1648. Gruijters, S.H.L.L., Veldkamp, J.G., Gunnink, J., Bosch, J.H.A., 2001. De lithologische en sedimentologische opbouw van de ondergrond van de Pannerdensche Kop. Eindrapport, NITG 01-166-B, TNO, the Netherlands. Hirano, M., 1972. Studies on variation and equilibrium state of a river bed composed of non-uniform material. Trans. Japn. Soc. Civil Eng. 4. Jansen, P.Ph., Van Bendegom, L., Van den Berg, J., De Vries, M., Zanen, A., 1979. Principles of River Engineering: The Non-Tidal Alluvial River. Pitman, London. Kalkwijk, J.P.Th., De Vriend, H.J., 1980. Computation of the ﬂow in shallow river bends. J. Hydraul. Res. IAHR 18 (4), 327–342. Kennedy, J.F., Nakato, T., Odgaard, A.J., 1984. Analysis, numerical modeling, and experimental investigation of ﬂow in river bends. In: Elliott, C.M. (Ed.), River Meandering, Proceedings of Conference on Rivers 1983, New Orleans, ASCE, 1984, pp. 843–856. Lesser, G.R., Roelvink, J.A., Van Kester, J.A.T.M., Stelling, G.S., 2004. Development and validation of a three-dimensional morphological model. Coastal Eng. 51 (8–9), 883–915. Meyer-Peter, E., Mu¨ller, R., 1948. Formulas for bed-load transport. Proceedings of second Congress IAHR, Stockholm, Paper No. 2, pp. 39–64. Mosselman, E., Hassan, K.I., Sieben, A., 2003. Effect of spatial grain size variations in two-dimensional morphological computations with uniform sediment. In: Sa´nchez-Arcilla, A. and Bateman, A. (Eds), Proceedings of IAHR Symposium on River, Coastal and Estuarine Morphodynamics, Barcelona, 1–5 September 2003, Publication of IAHR, Madrid, Spain, pp. 236–246. Mosselman, E., Sieben, A., Sloff, K., Wolters, A., 1999. Effect of spatial grain size variations on twodimensional river bed morphology. Procedings of IAHR Symposium on River, Coastal and Estuarine Morphodynamics, Genova, 6–10 Sept. 1999, Vol. I, pp. 499–507. Nelson, J.M., Smith, J.D., 1989. Evolution and stability of erodible channel beds. In: Ikeda, S. and Parker, G. (Eds), River meandering, AGU, Water Resources Monograph 12. pp. 321–377. Odgaard, A.J., 1981. Transverse bed slope in alluvial channel bends. J. Hydraul. Div. ASCE 107 (HY12), 1677–1694. Olesen, K.W., 1987. Bed topography in shallow river bends. PhD thesis, Delft University of Technology, Communications on Hydraulic and Geotechnical Engineering, No. 87-1, Delft University of Technology, ISSN 0169-6548. Paintal, A.S., 1971. Concept of critical shear stress in loose boundary open channels. J. Hydraul. Res. IAHR 9 (1), 91–108. Ribberink, J.S., 1987. Mathematical modelling of one-dimensional morphological changes in rivers with non-uniform sediment. PhD thesis, Delft University of Technology, Communications on Hydraulic and Geotechnical Engineering, No. 87-2, Delft University of Technology, ISSN 0169-6548. Shimizu, Y., Itakura, T., 1985. Practical computation of two-dimensional ﬂow and bed deformation in alluvial streams. Civil Engineering Research Report, Hokkaido Development Bureau, Sapporo. Sloff, C.J., Jagers, H.R.A., Kitamura, Y., Kitamura P., 2001. 2D Morphodynamic modelling with graded sediment. Proceedings of second IAHR Symposium on River, Coastal and Estuarine Morphodynamics, 10–14 September 2001, Obihiro, Japan, pp. 535–544. Struiksma, N., 1985. Prediction of 2-D bed topography in rivers. J. Hydraul. Engrg. ASCE 111 (8), 1169–1182. Struiksma, N., Olesen, K.W., Flokstra, C., De Vriend, H.J., 1985. Bed deformation in curved alluvial channels. J. Hydraul. Res. IAHR 23 (1), 57–79.

The importance of floods for bed topography and bed sediment composition

179

Van Bendegom, L., 1947. Some considerations on river morphology and river improvement. De Ingenieur 59, B1–B11 (in Dutch; English translation: National Research Council Canada, Technical Translation 1054, 1963). Wang, Z.B., Fokkink, R.J., De Vries, M., Langerak, A., 1995. Stability of river bifurcations in 1D morphodynamic models. J. Hydraul. Res. IAHR 33 (6), 739–750. Wilbers, A.W.E., Ten Brinke, W.B.M., 2003. The response of subaqueous dunes to ﬂoods in sand and gravel bed reaches of the Dutch Rhine. Sedimentology 50, 1013–1034.

Discussion by R. Ferguson You demonstrate that the main morphodynamic response to moderate ﬂoods is size sorting of the bed, but the main response to big ﬂoods is erosion and deposition. I was very interested in this since I too ﬁnd a combination of the two kinds of response in my own modelling work. You provide a neat explanation of your result in terms of a ratio of response times for bed composition and bed elevation, and show this reduces to the ratio of active layer thickness d to ﬂow depth h. This inspired me to think about the possible mechanisms and I have a suggestion which is entirely consistent with your results and explanation. Consider the fractional Exner equation in the form @ðð1 lÞdF i Þ @z (7:30Þ ¼ ð1 lÞ ðPi E i Þ @t @t where l denotes bed porosity and Fi, Pi, Ei denote the fraction of size i in the active layer, the bedload, and the exchange material, respectively, at the base of the active layer. If l and d are assumed constant over time, and Ei is equated with Fi as is traditional, the equation reduces to ð@F i =@tÞ Pi F i ¼ d ð@z=@tÞ

(7:31Þ

This conﬁrms that the rate of change of bed composition, relative to the rate of change of bed elevation, decreases with increased active layer thickness. It also quantiﬁes the tendency for composition change to be more important when transport is more size selective (bigger difference between Pi and Fi). For ﬁne fractions of the bed, Pi–Fi is positive and qFi/qt has the same sign as qz/qt; for coarse fractions the signs are reversed. As ﬂow depth and shear stress increase during major ﬂoods, transport becomes less size selective until Pi converges on Fi and no further change in bed composition is possible.

Reply by the authors We thank Rob Ferguson for the additional insights. His equation basically complies with our equation (7.22) under the assumption that there are no gradients in the sediment composition of the bedload ð@piT =@x ¼ 0Þ. This assumption, however, is responsible for his conclusion that no further change in bed composition is possible

The importance of floods for bed topography and bed sediment composition

179

Van Bendegom, L., 1947. Some considerations on river morphology and river improvement. De Ingenieur 59, B1–B11 (in Dutch; English translation: National Research Council Canada, Technical Translation 1054, 1963). Wang, Z.B., Fokkink, R.J., De Vries, M., Langerak, A., 1995. Stability of river bifurcations in 1D morphodynamic models. J. Hydraul. Res. IAHR 33 (6), 739–750. Wilbers, A.W.E., Ten Brinke, W.B.M., 2003. The response of subaqueous dunes to ﬂoods in sand and gravel bed reaches of the Dutch Rhine. Sedimentology 50, 1013–1034.

Discussion by R. Ferguson You demonstrate that the main morphodynamic response to moderate ﬂoods is size sorting of the bed, but the main response to big ﬂoods is erosion and deposition. I was very interested in this since I too ﬁnd a combination of the two kinds of response in my own modelling work. You provide a neat explanation of your result in terms of a ratio of response times for bed composition and bed elevation, and show this reduces to the ratio of active layer thickness d to ﬂow depth h. This inspired me to think about the possible mechanisms and I have a suggestion which is entirely consistent with your results and explanation. Consider the fractional Exner equation in the form @ðð1 lÞdF i Þ @z (7:30Þ ¼ ð1 lÞ ðPi E i Þ @t @t where l denotes bed porosity and Fi, Pi, Ei denote the fraction of size i in the active layer, the bedload, and the exchange material, respectively, at the base of the active layer. If l and d are assumed constant over time, and Ei is equated with Fi as is traditional, the equation reduces to ð@F i =@tÞ Pi F i ¼ d ð@z=@tÞ

(7:31Þ

This conﬁrms that the rate of change of bed composition, relative to the rate of change of bed elevation, decreases with increased active layer thickness. It also quantiﬁes the tendency for composition change to be more important when transport is more size selective (bigger difference between Pi and Fi). For ﬁne fractions of the bed, Pi–Fi is positive and qFi/qt has the same sign as qz/qt; for coarse fractions the signs are reversed. As ﬂow depth and shear stress increase during major ﬂoods, transport becomes less size selective until Pi converges on Fi and no further change in bed composition is possible.

Reply by the authors We thank Rob Ferguson for the additional insights. His equation basically complies with our equation (7.22) under the assumption that there are no gradients in the sediment composition of the bedload ð@piT =@x ¼ 0Þ. This assumption, however, is responsible for his conclusion that no further change in bed composition is possible

The importance of floods for bed topography and bed sediment composition

179

Van Bendegom, L., 1947. Some considerations on river morphology and river improvement. De Ingenieur 59, B1–B11 (in Dutch; English translation: National Research Council Canada, Technical Translation 1054, 1963). Wang, Z.B., Fokkink, R.J., De Vries, M., Langerak, A., 1995. Stability of river bifurcations in 1D morphodynamic models. J. Hydraul. Res. IAHR 33 (6), 739–750. Wilbers, A.W.E., Ten Brinke, W.B.M., 2003. The response of subaqueous dunes to ﬂoods in sand and gravel bed reaches of the Dutch Rhine. Sedimentology 50, 1013–1034.

Discussion by R. Ferguson You demonstrate that the main morphodynamic response to moderate ﬂoods is size sorting of the bed, but the main response to big ﬂoods is erosion and deposition. I was very interested in this since I too ﬁnd a combination of the two kinds of response in my own modelling work. You provide a neat explanation of your result in terms of a ratio of response times for bed composition and bed elevation, and show this reduces to the ratio of active layer thickness d to ﬂow depth h. This inspired me to think about the possible mechanisms and I have a suggestion which is entirely consistent with your results and explanation. Consider the fractional Exner equation in the form @ðð1 lÞdF i Þ @z (7:30Þ ¼ ð1 lÞ ðPi E i Þ @t @t where l denotes bed porosity and Fi, Pi, Ei denote the fraction of size i in the active layer, the bedload, and the exchange material, respectively, at the base of the active layer. If l and d are assumed constant over time, and Ei is equated with Fi as is traditional, the equation reduces to ð@F i =@tÞ Pi F i ¼ d ð@z=@tÞ

(7:31Þ

This conﬁrms that the rate of change of bed composition, relative to the rate of change of bed elevation, decreases with increased active layer thickness. It also quantiﬁes the tendency for composition change to be more important when transport is more size selective (bigger difference between Pi and Fi). For ﬁne fractions of the bed, Pi–Fi is positive and qFi/qt has the same sign as qz/qt; for coarse fractions the signs are reversed. As ﬂow depth and shear stress increase during major ﬂoods, transport becomes less size selective until Pi converges on Fi and no further change in bed composition is possible.

Reply by the authors We thank Rob Ferguson for the additional insights. His equation basically complies with our equation (7.22) under the assumption that there are no gradients in the sediment composition of the bedload ð@piT =@x ¼ 0Þ. This assumption, however, is responsible for his conclusion that no further change in bed composition is possible

180

E. Mosselman, K. Sloff

when the transport is no longer size selective. In principle, changes in bed composition remain possible if the composition of the bedload varies spatially. We do support Rob Ferguson’s remark that size selectiveness inﬂuences the tendency for composition change, but we believe the ratio of active-layer thickness to ﬂow depth to be the main inﬂuencing factor.

Gravel-Bed Rivers VI: From Process Understanding to River Restoration H. Habersack, H. Pie´gay, M. Rinaldi, Editors r 2008 Elsevier B.V. All rights reserved.

181

8 Review of effects of large ﬂoods in resistant-boundary channels Ellen Wohl

Abstract Gravel-bed channels formed in coarse sediment or mixed bedrock and coarse sediment have higher thresholds of resistance to ﬂuvial erosion than channels formed in ﬁnergrained sediment. These higher erosional thresholds are likely to be exceeded only during large ﬂoods. Variables that inﬂuence the geomorphic role of ﬂoods in resistantboundary channels include the ﬂood-generating mechanism (hydroclimatic and damburst processes), position within a drainage basin, erosional threshold, sediment supply, land use, in-channel wood, riparian vegetation, and time since last ﬂood. Geomorphic effects of ﬂoods include transport of sediment and wood, alteration of channel and valley morphology, and channel incision. Ecological effects of ﬂoods include alteration of physical and chemical characteristics of the river and ﬂoodplain, and changes in riparian and aquatic community composition and structure. Channel restoration and management must account for the occurrence and effects of infrequent, large magnitude ﬂoods by focusing on river process rather than river form. The net result of downstream trends in numerous physical variables is that the aggregate population of large ﬂoods creates the greatest geomorphic effects in headwater channels, whereas individual large ﬂoods are likely to be most geomorphically important in the middle portion of drainage basins.

1.

Introduction

This paper summarizes research on the geomorphic and ecological effects of large ﬂoods in resistant-boundary channels, and the implications of this research for channel management and restoration. An extensive literature now exists for the geomorphic and ecological impacts of ﬂoods in a variety of channel types, with resistant-boundary channels receiving a great deal of attention, including numerous books and review articles that cover at least some of the material addressed here (e.g., Baker et al., 1988; Beven and Carling, 1989; Tinkler and Wohl, 1998; Mosley, 2001). This review differs from previous work in that it addresses both the geomorphic and ecological roles of E-mail address: [email protected] ISSN: 0928-2025

DOI: 10.1016/S0928-2025(07)11125-1

182

E. Wohl

large ﬂoods in channels formed in either bedrock, coarse sediment, or some combination of the two substrates. In the context of this paper, effects refer to persistent features that are created during ﬂooding. A persistent feature is one that endures until the next ﬂood of similar magnitude occurs, as opposed to a transient feature that is altered by smaller ﬂows following the ﬂood (Brunsden and Thornes, 1979). Geomorphic ﬂood effects include erosional and depositional features. Ecological ﬂood effects include changes in species age distribution or composition in riparian and aquatic communities. Large floods have been deﬁned in a variety of ways for different ﬁeld settings, and the papers reviewed here use this phrase for ﬂows with widely differing recurrence intervals and differing ratios with respect to the mean annual ﬂood. In this paper, ﬂoods are referred to as being large if they are at least twice the magnitude of the mean annual ﬂood. Use of the word large in this context does not necessarily imply that the ﬂood will have geomorphic or ecological impacts that differ from those of the normal annual ﬂood in all ﬁeld settings; this depends very much on the details of channelboundary resistance to erosion relative to the hydraulic forces generated by the ﬂood. Resistant channel boundaries are streambeds and/or banks composed of either bedrock, alluvium that is coarser than pebble size (Z60 mm), or some combination of bedrock and coarse alluvium. These channels behave in ways distinctly different from channels formed in silt and sand because some minimum or threshold hydraulic force must be exceeded before substantial channel change begins in resistant-boundary channels, as opposed to the nearly constant adjustment of bedforms to changing hydraulic conditions in sand-bed channels (Simons and Richardson, 1966). Although sand-bed channels can have thresholds of motion (e.g., van Rijn, 1984 predicts a critical velocity of 0.3 m/s for a ﬂow depth of 1.5 m over medium sand), these thresholds are typically exceeded multiple times each year. In contrast, the much higher resistance thresholds of channels formed in coarser sediment or bedrock may be exceeded only a few times a year (e.g., Andrews, 1984), or once in decades or centuries (e.g., Grant et al., 1990; de Jong, 1994). Quantitative assessments of the geomorphic role of ﬂoods began with Wolman and Miller (1960), who proposed that the relative importance in geomorphic processes of extreme events can be measured in terms of (i) the relative amounts of work done on the landscape (which Wolman and Miller expressed as suspended sediment transport) and (ii) the formation of speciﬁc features of the landscape. Using sediment transport records from the United States, they concluded that in many basins the largest portion of the total sediment load is carried by ﬂows that recur on average once or twice each year, an analysis expanded in Leopold et al. (1964). Although the original Wolman and Miller paper discusses the increasing importance of infrequent events with increasing channel resistance to erosion, and a few studies focused on ﬁeld settings where extreme ﬂoods clearly dominated channel morphology or sediment transport (e.g., Tinkler, 1971; Baker, 1973), many subsequent investigators used the Wolman and Miller paper to develop a restrictive model of ﬂuvial dynamics that emphasized the importance of frequent ﬂows in most geomorphic settings. Nearly two decades elapsed before numerous ‘‘anomalous’’ case studies convinced the geomorphic community that some rivers are more likely to be dominated by ﬂoods of lower frequency. These rivers include channels with high seasonal and interannual ﬂow variability, high

Review of effects of large floods in resistant-boundary channels

183

ratios between discharge of infrequent ﬂoods and average annual ﬂow, abundant coarse bedload, high channel gradient, resistant channel boundaries, and highly erodible channel boundaries such as those found in braided rivers (Baker, 1977; Gupta, 1983; Nolan and Marron, 1985; Kochel, 1988; Baker and Kale, 1998; Cenderelli and Wohl, 2001, 2003; Wohl et al., 2001). Recent studies indicate that small to moderate ﬂoods can be very important for transporting sediment and shaping channel morphology even in gravel-bed and other resistant-boundary rivers (e.g., McKenney, 2001), but many resistant-boundary channels are dominated by large ﬂoods. This paper brieﬂy reviews the inﬂuences of ﬂood-generating mechanism, position in a drainage basin, erosional thresholds, sediment supply, land use, in-channel wood, riparian vegetation, and time elapsed since the last geomorphically effective ﬂood on the geomorphic role of ﬂoods. Land use, in-channel wood, and riparian vegetation indirectly inﬂuence the geomorphic role of ﬂoods by inﬂuencing erosional thresholds and sediment supply. They have separate subheadings in this paper because of their multiple roles. Examples of geomorphic and ecological impacts of ﬂoods are followed by discussions of the implications for channel management, and a synthesis of how the geomorphic role of ﬂoods varies across a drainage basin.

2. 2.1.

Variables inﬂuencing the geomorphic role of ﬂoods in resistant-boundary channels Flood-generating mechanism

Flood effects to resistant-boundary channels occur along a hydroclimatic spectrum in that climatic patterns creating ﬂood-generating precipitation differ in magnitude and frequency among different regions (e.g., Hirschboeck, 1987, 1988; Church, 1988; Gupta, 1988; Schick, 1988; Webb and Betancourt, 1990). For example, compilations of ﬂood records from the United States indicate that, for small-to moderate-sized basins less than approximately 30,000 km2, the largest values of peak discharge per unit drainage area occur in semiarid or arid regions (Costa, 1987), and in regions such as Hawaii where mountainous islands intercept moisture from convective systems and tropical storms (O’Connor and Costa, 2004). Focusing speciﬁcally on the western United States, Pitlick (1994) found that large ﬂoods where streamﬂow is dominated by snowmelt, frontal rainfall, or rain-on-snow can be two to four times the average annual ﬂood and recur every couple decades, whereas large ﬂoods in drier regions dominated by convective rainfall can be more than ten times the average annual ﬂood and recur only a few times each century. The much greater interannual variability of ﬂood magnitude in drier regions and in some parts of the seasonal tropics means that erosional and depositional features created during large ﬂoods can persist during subsequent smaller ﬂows either because the features exist in portions of the channel that are not submerged during smaller ﬂows, or because the smaller ﬂows do not exceed the erosional thresholds necessary to modify cohesive or very coarse-grained channel boundaries. These types of persistent ﬂood effects have been described for gravel-bed and bedrock rivers in the seasonal

E. Wohl

184

tropics (Gupta and Dutt, 1989; Wohl, 1992a, b; Kale and Hire, 2004) and semiarid/ arid regions (Patton and Baker, 1976; Graf, 1988; Tooth, 2000). Climatic inﬂuences on ﬂooding are also exacerbated by topography. The largest ﬂood ﬂows generally result from orographic enhancement of precipitation (Shroba et al., 1979; SturdevantRees et al., 2001; O’Connor et al., 2002). Topographic variability can also alter storm tracks and promote rapid runoff that enhances ﬂooding (O’Connor and Costa, 2004). Substantial geomorphic change can result from ﬂoods that are not directly produced by precipitation, including ﬂoods from breached natural (landslide, moraine, ice) or artiﬁcial dams. These can be the most geomorphically effective ﬂoods within a channel network because they are capable of generating a discharge two or three orders of magnitude greater than meteorologically generated discharges for the drainage basin. Damburst and outburst ﬂoods are likely to be particularly geomorphically important in high relief valleys with unstable hillslopes that create tall landslide dams (e.g., Tianche et al., 1986; Hewitt, 1998; Korup, 2005), and in drainages that still have active glaciers (e.g., Maizels, 1991, 1997; Gomez et al., 2002). Numerous compilations (e.g., Costa and Schuster, 1988; Shroder et al., 1998; Cenderelli, 2000; O’Connor et al., 2002) and case studies (e.g., Baker, 1973; Carling and Glaister, 1987; Desloges and Church, 1992; O’Connor, 1993; Benito et al., 1998; Cenderelli and Wohl, 2003) describe the geomorphic effects of damburst and outburst ﬂoods in resistant-boundary channels.

2.2.

Position within a drainage basin

Several studies indicate that the relative geomorphic importance of ﬂoods varies with position in a drainage basin (e.g., Clark et al., 1987; Miller, 1990). Smaller sub-basins in the headwater portion of a drainage are more likely to have the resistant channel boundaries, abundant coarse bedload, and high channel gradient that promote major geomorphic response to ﬂooding (Patton, 1988; Jacobson et al., 1989; Kale and Hire, 2004). These channels are frequently disturbed because their smaller drainage area often results in a high percentage of contributing area, and because proximity to adjacent hillslopes and lack of ﬂoodplain and valley bottom storage minimizes time for precipitation to concentrate in channels (Clark et al., 1987). Because geomorphic response to individual large ﬂoods decreases with decreasing time interval between large ﬂoods (Harvey, 1984; Cenderelli and Wohl, 2003), the geomorphic effectiveness of any single large ﬂood is likely to be less in the headwaters than in other portions of the drainage. Large ﬂoods in the aggregate, however, are likely to dominate the headwater channels. The steepest, narrowest channel segments in headwater areas are also subject to disturbance from mass movements such as debris ﬂows that originate on adjacent hillslopes (e.g., Froehlich and Starkel, 1987; Larsen and Roman, 2001; Korup, 2005). The relationship between ﬂoods and debris ﬂows is spatially and temporally complex. Tributary debris ﬂows can be a major source of sediment and wood mobilized during ﬂoods (Webb et al., 1989; Wohl and Pearthree, 1991; Benda and Dunne, 1997; Benda et al., 2003; Montgomery et al., 2003); debris ﬂows can alternate with ﬂoods along the main channel over timespans of months to years (Seidl and Dietrich, 1992; Stock

Review of effects of large floods in resistant-boundary channels

185

and Dietrich, 2003; Bunn and Montgomery, 2004); and individual large ﬂows can alternate downstream between ﬂoods, hyperconcentrated ﬂows, and debris ﬂows as sediment concentration varies (Cenderelli and Wohl, 1998; O’Connor et al., 2001). As drainage area increases downstream, the probability that the entire basin will be contributing to a ﬂood decreases. Storms capable of generating large ﬂoods are less common than in the headwaters portion of the drainage and frequency of large ﬂoods decreases. The attenuation of water and sediment discharges associated with ﬂoodplains also increases, further reducing the likelihood of large ﬂoods. The long intervals between large ﬂoods increases the likelihood that normal ﬂows will transport volumes of sediment comparable to those transported by the large ﬂoods, as well as substantially reworking geomorphic features created during large ﬂoods. Working in southern New England, for example, Patton (1988) found that ﬂoodplain erosion and deposition during rare large ﬂoods was much less substantial in lowland drainages than in smaller, steeper highland drainages. As a result of downstream decreases in large ﬂood recurrence interval and magnitude relative to annual ﬂoods, individual large ﬂoods can be most geomorphically effective in the middle portions of drainages where such ﬂoods are less frequent than in the headwaters, but still exceed erosional thresholds sufﬁciently to cause substantial sediment transport and channel change (Clark et al., 1987). Flooding in the central Appalachian Mountains of the United States associated with a dissipating tropical cyclone in 1985 provides an example. Prolonged and locally intense precipitation produced over 250 mm of rain during 3 days, and this followed an unusually wet month (Clark et al., 1987). The resulting ﬂoods exceeded the 100year discharge in some drainage basins and the 500-year discharge in other basins. Clark et al. (1987) found that basins draining 130–4,000 km2 were most geomorphically altered by the ﬂood. Smaller basins receive comparable precipitation intensities during summer thunderstorms that occur much more frequently than the 1985 storm, and larger basins included sub-basins that received much less precipitation during the 1985 storm.

2.3.

Erosional threshold

The importance of ﬂoods in individual resistant-boundary channels will also depend on the frequency/duration of ﬂows exceeding an erosional threshold governed by channel-boundary composition (Costa and O’Connor, 1995) (Fig. 8.1). Following Bull’s (1979) choice of stream power as an appropriate descriptor of threshold conditions for channel change, Magilligan (1992) and Wohl et al. (2001) proposed minimum values of stream power necessary to initiate substantial geomorphic change in alluvial and bedrock channels, respectively. Alternatively, Miller (1990) concluded that unit stream power alone is not a reliable predictor of geomorphic change for individual sites because of the complex interactions among channel width and gradient, channel pattern, spatial arrangements of roughness elements, and local ﬂow obstructions. The geomorphic effectiveness of a ﬂood will ultimately depend on the duration of ﬂow(s) that exceed the erosional resistance of the channel boundaries. Erosional resistance is governed by bedrock and alluvial characteristics (Table 8.1). The ability

E. Wohl

186

a

b

Unit stream power (W/m2)

bedrock erosion threshold

energy available for geomorphic change

alluvial erosion threshold

c

Time Figure 8.1. Conceptual diagram of energy available for geomorphic change during ﬂoods as a function of ﬂood magnitude and duration above thresholds for erosion of different types of channel substrate (see Costa and O’Connor, 1995, Fig. 11). The three curves represent three different types of idealized ﬂood hydrographs: (a) could be a convective storm over a small basin, (b) could be a frontal rainfall storm, and (c) could be a seasonal snowmelt peak of longer duration but lower magnitude than (a) and (b).

of ﬂow to exceed threshold resistance is not simply a matter of average shear stress. Fluctuations about the mean associated with turbulence can be more important in initiating bedrock erosion or particle entrainment that, once started, promotes a selfenhancing feedback by inducing further ﬂow separation and turbulence (Nelson et al., 1995; Robert et al., 1996; Lawless and Robert, 2001). It has proven extremely difﬁcult to adequately quantify the factors inﬂuencing boundary erosional resistance and hydraulic driving forces because these factors vary across temporal and spatial scales. For example, many resistant-boundary channels have both bedrock and alluvium exposed along the channel boundaries (Wohl, 1998). An effective approach is likely to be a probabilistic characterization of the key control variables (Powell, 1998; Graf, 2001), but hydraulic models that simulate the two- or three-dimensional ﬂow properties necessary for such a probabilistic approach are just now becoming available to the research community (Miller and Cluer, 1998; Booker et al., 2001; Nicholas, 2001).

2.4.

Sediment supply

Sediment supply exerts a critical control on the type and magnitude of erosional and depositional features produced by a ﬂood. Erosion of bedrock channels is greatest at moderate sediment supplies (Seidl et al., 1994; Sklar and Dietrich, 2004); for example, smaller amounts of sediment do not provide sufﬁcient tools for abrasion of channel boundaries, and larger amounts of sediment mantle the channel bed and protect it

Review of effects of large floods in resistant-boundary channels Table 8.1.

187

Factors inﬂuencing erosional resistance in bedrock and gravel-bed channels.

Substrate characteristic

Associated erosional processes

Sample reference

Joint geometry (orientation, spacing, width, continuity) and analogous discontinuities (e.g., bedding planes) [bedrock] Porosity and permeability [bedrock]

Cavitation initiated along joints; quarrying lifts separated blocks

Baker and Pickup (1987), Miller (1991), Tinkler (1993), Hancock et al. (1998), Wohl and Springer (2005) Howard (1998), Wohl et al. (1999)

Crystal or grain boundaries [bedrock] Sediment cover [bedrock]

Clast size, shape, and sorting (mean size, sorting, particle shape, packing) [alluvial] Clast resistance [alluvial]

Bank stratigraphy [alluvial]

Channel and valley geometry [bedrock and alluvial]

Seepage and enhanced chemical weathering can produce surface irregularities that induce ﬂow separation and accelerated abrasion Differential weathering produces microscale surface roughness that induces ﬂow separation and accelerated abrasion Areal extent and thickness of alluvial veneer in bedrock channels determines effectiveness of sediment in either shielding bedrock surface from erosion, or promoting erosion through abrasion These characteristics inﬂuence entrainment threshold for drag and lift forces Resistance of clasts to chemical and mechanical weathering determines relative importance of disintegration in place versus downstream transport Inﬂuences importance of particle-by-particle erosion versus bank slumping or other forms of collapse Inﬂuences distribution of hydraulic variables by governing boundary roughness and generation of turbulence, which in turn inﬂuences cavitation, lift, and abrasion

Seidl et al. (1994), Sklar and Dietrich (1998, 2001)

Kirchner et al. (1990), Powell (1998), Shvidchenko et al. (2001) Kodama (1994), Sklar and Dietrich (2001)

Lawler (1992), Fonstad and Marcus (2003) Shroba et al. (1979), Miller and Parkinson (1993), Wohl et al. (2001)

from erosion. At the most general level, bed coarsening (e.g., Dietrich et al., 1989), or channel erosion result when transport capacity exceeds sediment supply during a ﬂood in a predominantly alluvial channel. Deposition, which is usually localized in resistant-boundary channels, results when sediment supply exceeds transport capacity during a ﬂood (e.g., Nolan and Marron, 1985; Wohl, 1992a). Sediment supply during a ﬂood can reﬂect hillslope, valley bottom, and in-channel processes. Hillslope sediment inputs result primarily from slope instability triggered

E. Wohl

188

by surface runoff, increased inﬁltration, or undercutting of the slope by ﬂoodwaters (e.g., Williams and Guy, 1973). Valley bottom sediment can be mobilized by overbank ﬂows that exceed the threshold shear stress for erosion of the ﬂoodplain and valley bottom, which depends on characteristics including grain size, vegetation, and land use. In-channel sediment supply resulting from bed and bank erosion, like the other sources of sediment supplied to ﬂoods, can be highly variable in time and space. The presence of a coarse surface layer is likely to be particularly important in controlling in-channel sediment supply in resistant-boundary channels. This coarse surface layer (which has been variously referred to as an armor, pavement, or censored layer; Carling and Reader, 1982) can reduce the sizes and amount of sediment transported (e.g., Gomez, 1983; Parker and Sutherland, 1990). The presence of a coarse surface layer does not imply that the bed is static. Numerous investigators have demonstrated transport of a majority of the particle sizes present on the bed while the coarse surface layer remains stable (e.g., Parker and Klingeman, 1982; Andrews and Erman, 1986). Powell (1998) and Parker and Toro-Escobar (2002) provide more detailed reviews of the mechanics of coarse surface layers and their effects on sediment transport.

2.5.

Land use

Contemporary or historical land use can strongly inﬂuence erosional thresholds and sediment during a ﬂood. Channel erosional thresholds can be altered by channel stabilization techniques such as riprap (Kresan, 1988), by the location of structures such as dams, bridges, or road crossings (Anthony and Julian, 1999; Chin and Gregory, 2001), or by alteration of channel geometry and channel–ﬂoodplain connections as a result of dredging, channelization, or construction of levees (e.g., Brookes, 1988; Wyzga, 1996; Wohl, 2000a). Land uses such as timber harvest (e.g., Madej and Ozaki, 1996; Stover and Montgomery, 2001), agriculture (e.g., Klimek, 1987; Starkel, 1988; Mei-e and Xianmo, 1994; Clark and Wilcock, 2000), roads (e.g., Froehlich, 1991), or mining (e.g., Macklin et al., 1992), can result in large pulses of sediment to adjacent stream channels during periods of rainfall-induced ﬂooding. Laboratory experiments (Lisle et al., 1997; Cui et al., 2003a), numerical simulations (Cui et al., 2003b), and ﬁeld studies (Madej and Ozaki, 1996; Madej, 2001; Sutherland et al., 2002) suggest that these pulses move by translation and dispersion over time intervals dependent on ﬂow magnitude and duration following introduction of the sediment pulse. Sediment pulses can alter the grain-size distribution, bedforms, and planform of the channel as the pulses move downstream (e.g., Knighton, 1989; Hilmes and Wohl, 1995; James, 1999). Conversely, urbanization generally reduces sediment supply during ﬂoods as a result of increased paved areas, and this commonly results in channel erosion (e.g., Wolman, 1967; Roberts, 1989; Booth, 1990). Dams and other forms of ﬂow regulation also inﬂuence channel response to ﬂoods by changing the characteristics of both water and sediment discharge (e.g., Young et al., 2001; Grant et al., 2003). The details of these changes depend on the type of ﬂow regulation. Dams generally reduce peak ﬂows and increase base ﬂows, but the

Review of effects of large floods in resistant-boundary channels

189

magnitude of these effects varies markedly among individual dams (Graf, 1999). The most common scenario on dammed rivers is that the geomorphic role of moderate to large ﬂoods is substantially reduced because of reductions in the magnitude and frequency of these ﬂoods (e.g., Williams and Wolman, 1985; Collier et al., 1996; Magilligan et al., 2003). By trapping bedload and at least a portion of suspended load, dams can also cause downstream coarsening, bed incision, bank erosion, and change in channel planform (e.g., Collier et al., 1996; Kondolf, 1997).

2.6.

In-channel wood

The presence of wood along the channel can exert a substantial control on the location, volume, and stability of stored sediment (Lancaster et al., 2001; Abbe and Montgomery, 2003; Bunn and Montgomery, 2004). Wood jams in large streams can create nucleation sites for bars, and in some cases promote a multi-thread channel (Collins and Montgomery, 2001; O’Connor et al., 2003). Pieces of wood in smaller streams can be longer than the channel width and/or partially attached to the channel bank, and thus provide very effective points of stability for trapping sediment until the wood rots (Keller and Tally, 1979). This sediment trapping alters bedforms; creates bars, increases step height (Keller and Swanson, 1979; Thompson, 1995; Curran and Wohl, 2003; Gomi et al., 2003; MacFarlane and Wohl, 2003); reduces channel gradient (Faustini and Jones, 2003); promotes substrate heterogeneity and local ﬁning (Kail, 2003); and creates alluvial reaches where bedrock would otherwise form the channel substrate (Montgomery et al., 1996). Wood also deﬂects ﬂow and creates localized scour that increases pool volume and bank undercutting (Fausch and Northcote, 1992; Baillie and Davies, 2002), as well as promotes overbank ﬂooding (Jeffries et al., 2003). Removal of wood from smaller channels commonly results in greater sediment ﬂuxes as local sediment deposits are mobilized (Heede, 1985; Klein et al., 1987; Smith et al., 1993), and can trigger widespread channel instability in the form of bed and bank erosion (Brooks et al., 2003). When comparing forested rivers around the world, wood may not play as important a geomorphic role on European rivers with a long history of forest and channel management that effectively reduces the size and volume of wood recruited to stream channels (Pie´gay and Gurnell, 1997; Pie´gay et al., 1999).

2.7.

Riparian vegetation

Riparian vegetation can mediate channel response to ﬂooding by increasing bank resistance through roots (e.g., Gray and Barker, 2004; Pollen et al., 2004; Rutherfurd and Grove, 2004; Pollen and Simon, 2005) and hydraulic roughness of overbank areas (e.g., Pie´gay and Bravard, 1997; Kean and Smith, 2004). Along some steep, coarse-grained streams, riparian vegetation may be critical in maintaining a singlethread channel with adjacent ﬂoodplains, rather than a braided planform extending across much of the valley bottom (Smith, 2004). Discontinuous vegetation may also direct ﬂow toward the opposite bank and promote meandering (Bennett et al., 2002).

E. Wohl

190

Along multi-thread channels, vegetation reduces total channel width, braiding index, and relative mobility of channels (Tal et al., 2004). Vegetated areas between sub-channels in multi-thread channel systems can create threshold effects such that smaller discharges conﬁned between vegetated areas are primarily erosive, whereas larger discharges that partially inundate hydraulically rough vegetated areas are primarily depositional (Merritt and Wohl, 2003). Vegetation helps to stabilize the interﬂuve areas and promote the formation of sub-channels that reduce ﬂow resistance as ﬂow depth increases (Wende and Nanson, 1998). In cases where vegetation establishment is facilitated by regulation-induced changes in ﬂow regime, channels have changed from a braided to a meandering or anastomosing planform (Nadler and Schumm, 1981; Eschner et al., 1983; Johnson, 1994; Pie´gay and Salvador, 1997). 2.8.

Time since last flood

Another important component of ﬂow duration above an erosional threshold is the time elapsed since the last geomorphically effective ﬂood. Beven (1981) used hypothetical scenarios to demonstrate that geomorphic effectiveness should depend on the ordering of ﬂoods of differing magnitude, the time elapsed between subsequent ﬂoods, and the state of the channel at the time of a particular ﬂood, such that real channels are most appropriately considered to have a time variable, rather than a ﬁxed, threshold of erosion. Several investigators have found that when a channel is modiﬁed by a large ﬂood, subsequent large ﬂoods cause very little geomorphic change if intervening lower ﬂows did not modify erosional and depositional features created by the ﬁrst large ﬂood (Harvey, 1984; Cenderelli and Wohl, 2003). The time elapsed since the last large ﬂood that destabilized a resistant-boundary channel and exposed new sources of sediment can also determine sediment supply. For example, the ratio of bedload transport volume to effective runoff increased substantially during the years following a 50-year ﬂood in the Rio Cordon of Italy (Lenzi et al., 2004).

3. 3.1.

Geomorphic effects of ﬂoods Sediment dynamics

Both suspended and bedload transport can be orders of magnitude higher during a large ﬂood than during average ﬂows if the large ﬂood exceeds the erosional threshold set by a coarse surface layer in gravel-bed channels (Inbar, 1987; Eaton and Lapointe, 2001). High rates of sediment transport can trigger channel-boundary erosion, resulting in destabilization of adjacent hillslopes (Scott and Gravlee, 1968; Burbank et al., 1996), change of local base level for upstream or tributary channel segments (Sloan et al., 2001), or loss of ﬂoodplain area. Increased sediment movement during ﬂoods can change a step-pool or pool-rifﬂe channel to plane-bed morphology, although the pre-ﬂood bed morphology generally gradually reappears over a period of months to years following the ﬂood (Lisle, 1982; Sawada et al., 1983; Wohl and Cenderelli, 2000; Lenzi, 2001; Madej, 2001). The forces exerted by large masses of

Review of effects of large floods in resistant-boundary channels

191

bedload can break or dislodge the in-channel wood that stabilizes high-gradient gravel-bed rivers and creates sediment storage sites along both high- and low-gradient gravel-bed channels (Keller and Swanson, 1979; Smith et al., 1993; Montgomery et al., 1995; Carling and Tinkler, 1998; Bufﬁngton and Montgomery, 1999; Bufﬁngton et al., 2002; O’Connor et al., 2003; Bunn and Montgomery, 2004). The great majority of studies measuring sediment transport in resistant-boundary channels during ﬂoods indicate that bedload transport in particular is highly variable in time and space, and not adequately modeled by transport equations developed for channels formed in ﬁner-grained alluvium (e.g., Carson and Grifﬁths, 1987; Gomez and Church, 1989; Wohl, 2000b). Quantitative descriptions and predictions of sediment dynamics during ﬂoods in resistant-boundary channels remains an area where much more work is needed.

3.2.

Erosional and depositional features

Sediment supply interacts with channel and valley geometry to control erosional and depositional features created during ﬂoods along resistant-boundary channels. Floodplain erosional features include longitudinal grooves, stripping of alluvium, and secondary anastomosing channels (Miller and Parkinson, 1993). Flood-induced erosional features within cohesive-boundary channels include longitudinal grooves, potholes, inner channels, and various types of abraded facets (Baker and Pickup, 1987; Wohl, 1993; Hancock et al., 1998; Springer and Wohl, 2002). Erosion within non-cohesive channel boundaries can result in channel widening or deepening, coarsening of grain-size distribution on the streambed, or change in bedform conﬁguration (Stewart and LaMarche, 1967; Williams and Guy, 1973; Miller and Parkinson, 1993; Wondzell and Swanson, 1999). Flood depositional features derive from ﬁne sediments carried in suspension that create slackwater deposits (Baker, 1987; Baker and Pickup, 1987; Kite and Linton, 1993) and various types of channel-margin sandbars such as separation deposits or reattachment deposits (Schmidt, 1990; Schmidt and Rubin, 1995; Cenderelli and Cluer, 1998). Depositional features created from coarser sediment carried as bedload include expansion bars downstream from constricted reaches (Baker, 1978, 1984; O’Connor, 1993); longitudinal bars along valley margins at local ﬂow expansions (Stewart and LaMarche, 1967; Carling, 1987; Zielinski, 2003); point bars along the inner margins of valley bends (de Jong and Ergenzinger, 1995); pendant bars downstream from obstructions (Scott and Gravlee, 1968); imbricate clusters upstream from obstructions (Cenderelli and Cluer, 1998); and, where ﬂoodwaters break through natural or artiﬁcial levees, gravel bars, sand sheets, and splay deposits on ﬂoodplains (Kite and Linton, 1993; Miller and Parkinson, 1993). As might be expected, a ﬂood along any given channel is more likely to be erosional in steeper, narrower channel reaches and depositional in lower gradient, wider reaches (Stewart and LaMarche, 1967; Malde, 1968; Shroba et al., 1979; Cenderelli and Wohl, 2003). Because large ﬂoods generate extreme hydraulic forces capable of creating erosional and depositional features that subsequent lower discharges cannot substantially modify, some ﬂood-induced forms can be used to estimate the magnitude of

E. Wohl

192

paleodischarges along a channel (O’Connor et al., 1986; Baker, 1987; Nott and Price, 1994; Zen and Prestegaard, 1994; Jansen and Brierley, 2004). The higher the erosional resistance of the channel boundary, the more likely that ﬂood-induced forms will persist over time intervals equal to or greater than the recurrence interval of the ﬂood magnitude that created the forms. 3.3.

Channel planform

Floods can also alter channel planform, most commonly creating a braided pattern along what was previously a straight or meandering channel (Warburton, 1994; Friedman et al., 1996), or alter channel location through processes such as avulsion (Lapointe et al., 1998). Channel morphology (bedforms, channel pattern), surface and subsurface grain-size distributions, stream size, and channel constraint all inﬂuence hyporheic exchange ﬂows along high-gradient streams (Kasahara and Wondzell, 2003). Floods can alter hyporheic exchanges between instream and subsurface ﬂow by altering channel morphology and grain-size distributions. Flood effects along fourthand ﬁfth-order channels in the Lookout Creek catchment of Oregon, for example, varied from large changes in unconstrained stream reaches where the channel incised during the ﬂood, to lesser changes in constrained stream reaches where bedrock boundaries limited the depth and area of sediment available to be reworked by the ﬂood (Wondzell and Swanson, 1999).

4.

Ecological effects of ﬂoods

Ecological effects of ﬂoods include disturbance of aquatic and riparian communities. A disturbance is deﬁned in this context as a relatively discrete event in time that disrupts ecosystem, community, or population structure, and that changes resources, availability of substratum, or the physical environment (Pickett and White, 1985). Flood-induced disturbances alter river and ﬂoodplain characteristics as diverse as instream water temperature and chemistry (e.g., Brunke and Gonser, 1997); lateral and longitudinal nutrient ﬂows (Fisher et al., 1998); streambed grain size, heterogeneity, and stability (e.g., Erman et al., 1988); bedform type and dimensions (e.g., Fausch and Bramblett, 1991); hyporheic exchange patterns (Wroblicky et al., 1998); channel-margin germination sites for plants (e.g., Swanson et al., 1998); and ﬂoodplain habitat availability (Junk et al., 1989). Flood-related processes are essential to the existence of many aquatic and riparian organisms because the processes create and maintain habitat, and because they present recruitment opportunities for new organisms (e.g., riparian seedling germination) and species. Although the annual ﬂood is not normally considered a disturbance (Sparks et al., 1990), this scale of ﬂooding provides cues that species rely on for timing migration, spawning or seed dispersal, hatching or germination, and seasonal growth (Junk et al., 1989; Nilsson et al., 1991; Power et al., 1995; Merritt and Wohl, 2002). The annual ﬂood also facilitates lateral, longitudinal, and vertical exchanges of nutrients, organic matter, and organisms, and increasingly provides a necessary diluting ﬂow

Review of effects of large floods in resistant-boundary channels

193

that disperses contaminants throughout the river ecosystem (Sparks et al., 1990; Ciszewski, 2001). Numerous adverse ecological impacts can thus result from changes in the magnitude or duration of the annual ﬂood. Rarer and larger ﬂoods, by creating the quasi-permanent erosional and depositional features described earlier, structure the riverine environment in ways that govern the distribution of organisms (Fausch et al., 2002). For example, Everitt (1968) demonstrated that the lateral and longitudinal distribution of bands of even-aged cottonwood (Populus sargentii) along the Little Missouri River in North Dakota reﬂected ﬂood history and associated seedling germination sites. Flow regulation and loss of ﬂood peaks along analogous channels has resulted in substantial changes in species composition and community structure of riparian vegetation (Scott et al., 1996). Riparian vegetation is structured by the frequency, duration, and intensity of ﬂoods, as well as availability and stability of germination sites, supply of propagules, and water availability (Birkeland, 1996; Hupp and Osterkamp, 1996; Pie´gay and Bravard, 1997). Changes in any of these characteristics will change riparian vegetation communities (Nilsson and Jansson, 1995; Nakamura and Shin, 2001; Johansson and Nilsson, 2002). Similarly, changes in ﬂood regime will alter instream characteristics including grain-size distribution, hyporheic exchange, nutrient retention, wood loading, pool volume, channel cross-sectional and planform geometry, bank stability, and habitat abundance and diversity, all of which are crucial to the survival of aquatic organisms such as macroinvertebrates and ﬁsh (McKenney, 2001; Pitlick and Wilcock, 2001).

5.

Implications of ﬂood effectiveness for channel management and restoration

In a widely cited paper, Poff et al. (1997) emphasize the importance of the natural ﬂow regime in a river as a ‘‘master variable’’ that limits the distribution and abundance of riverine species and regulates ecological integrity by inﬂuencing water quality, energy sources, physical habitat, and biotic interactions. They list ﬁve basic components of the ﬂow regime – magnitude, frequency, duration, timing, and rate of change – that must be preserved in order to conserve ecological integrity. Geomorphologists also consider the sediment regime to be a primary determinant of channel morphology and process, particularly because alterations in sediment supply, as well as ﬂow regime, have substantially changed many human-impacted rivers (e.g., Sear, 1994; Pitlick and Wilcock, 2001). The Colorado River in the Grand Canyon, for example, continues to have coarse sediment supplied by episodic debris ﬂows on tributaries downstream from Glen Canyon Dam (Webb et al., 1989). The trapping of most ﬁner, suspended sediment behind the dam, however, and the loss of large ﬂood peaks that historically deposited this ﬁne sediment high along the channel margins, has led to erosion of beaches and terraces that provide important riparian habitat, as well as cultural and recreational sites (Collier et al., 1996). What emerges from review of the literature is that ﬂoods are integral to physical and biological conditions along river corridors (Ligon et al., 1995; Graf, 2001; Bunn and Arthington, 2002). Changing ﬂood magnitude and duration affects riverine species directly, through the loss of nutrients, habitat, and migration or dispersal corridors, and indirectly, through the loss of habitat-maintaining processes such as

E. Wohl

194

ﬂushing of ﬁne sediment from spawning gravels (Harvey et al., 1993), pools (Wohl and Cenderelli, 2000), and channel-margin bars (Auble et al., 1994; Elliott and Parker, 1997). Numerous studies indicate that some aspects of channel adjustment require ﬂoods in order to be maintained (Table 8.2). The magnitude and frequency of these formative ﬂoods varies between individual rivers and speciﬁc channel features, largely as a function of boundary resistance. Along many rivers, the annual ﬂood can keep spawning gravels and pools ﬂushed of ﬁne sediment, for example, but less frequent large ﬂoods are necessary to redistribute hillslope-derived coarse sediment

Table 8.2.

Examples of channel adjustment dependent on ﬂoods.

Location

Description

References

Black Canyon of the Gunnison River, Colorado, USA

Deep, narrow gorge with dam upstream; continuing sediment inputs from valley walls and tributaries; lack of ﬂood ﬂows to ﬂush ﬁne sediment from canyon and redistribute coarse sediment along river Canyon with dam upstream; continuing sediment inputs from tributaries; lack of ﬂood ﬂows to bring ﬁne sediment into suspension and deposit it high on the channel margins, and redistribute coarse sediment along river River in wide canyon; simultaneous introduction of exotic riparian plant tamarisk (Tamarix spp.), regulation of ﬂow from dam upstream, and natural reduction in ﬂood magnitudes caused channel narrowing Only large, rare ﬂoods are capable of mobilizing coarse sediment forming gravel bars and rifﬂes/rapids in these canyon systems Only large, rare ﬂoods are capable of mobilizing very large clasts that anchor steppool sequences Periodic large ﬂoods reverse process of channel narrowing to a single channel, creating a braided pattern

Liquori (1995), Elliott and Parker (1997), Elliott and Hammack (2000), Dubinski, 2005

Grand Canyon of the Colorado River, Arizona, USA

Green River, Utah, USA

Boulder Creek, Utah, USA Burdekin River, Australia Tapi River, India Step-pool channel in British Columbia, Canada Gila River, Arizona, USA Glacial stream, Switzerland Plum Creek, Colorado, USA

Kieffer (1985), Collier et al. (1996), Wiele et al. (1996), Rubin et al., 1998

Graf (1978), Allred and Schmidt (1999), Merritt and Cooper, 2000

O’Connor et al. (1986), Wohl (1992a), Kale and Hire, 2004 Zimmermann and Church, 2001 Burkham (1972), Warburton (1994), Friedman et al., 1996

Review of effects of large floods in resistant-boundary channels

195

Figure 8.2. Talus deposit impinging on the channel of the Gunnison River in the Black Canyon of the Gunnison, Colorado. Flow is from left to right. These coarse sediments introduced directly to the channel from adjacent valley walls can only be mobilized by ﬂows that recurred on average once every 2–3 years (the ﬂow has been regulated since 1966, increasing the recurrence interval of these ﬂows to 40 years). Photograph courtesy of Ian Dubinski.

and maintain the downstream spacing of pools and rifﬂes (O’Connor et al., 1986; Wohl, 1992a; Dubinski, 2005) (Fig. 8.2). The primary implication of the physical and ecological importance of ﬂoods on a wide range of river types is that rivers must be managed and restored for process rather than form, and process includes ﬂoods (Stanford et al., 1996). Restoring form without process by engineering a speciﬁc channel conﬁguration that cannot be maintained over a period of decades or longer by the existing ﬂow regime, for example, is likely to require continual, expensive artiﬁcial maintenance and is unlikely to replicate the conditions necessary for a fully functional river ecosystem (Kondolf et al., 2001; Wohl et al., 2005). Of equal importance to river management is recognition that ﬂood prevention is a difﬁcult matter. Perceived detrimental effects of ﬂoods have been mitigated by levees, reservoirs, and ﬂood-detention basins, and other structural, warning, and zoning measures, but such efforts generally had only limited and partial success. Even smaller, apparently thoroughly controlled mountain streams that are extensively channelized and stepped over grade-control structures can, and will, experience unexpected large ﬂoods that remove these structures and thoroughly rework the channel boundaries (e.g., Gavrilovic and Matovic, 1991). Given the inevitability of ﬂoods, the most effective management strategies will be those that restore and maintain a nearly natural ﬂow regime and sediment supply, and permit the river to adjust to ﬂuctuations in water and sediment discharge within a riverine corridor that has the minimum possible structural constraints (Kondolf, 1996; Graf, 2001; Ward et al., 2001; Jaquette et al., 2005).

E. Wohl

196

Communicating the geomorphic and ecological importance of large ﬂoods to resource managers and the public remains a difﬁcult task. Differentiating the role of large ﬂoods with respect to position in the drainage basin, as well as speciﬁc ﬂood effects (e.g., winnowing ﬁne sediments from spawning gravels), may facilitate communication.

6.

The geomorphic role of ﬂoods revisited

Wolman and Miller’s classic 1960 paper assumed that suspended sediment represented the greatest volume of material moved in most rivers, and that therefore the basic geomorphic role of rivers – that of gradually transporting sediment downstream – was most appropriately characterized in terms of the magnitude of ﬂow that transported the most suspended sediment over a period of many years. As subsequent work has expanded concepts of the geomorphic role of ﬂoods to include the persistence of ﬂoodcreated landforms such as channel pattern, bedforms, and grain-size distribution, it has become more difﬁcult to agree on a consistent, readily applicable method for quantifying the geomorphic role of any particular ﬂow magnitude. Wolman and Miller (1960) focused on alluvial systems. Subsequent studies, as outlined in this review, gave more attention to resistant-boundary channels where, by deﬁnition, thresholds of changes are higher, more variable, and thus more challenging to deﬁne. As summarized previously, the geomorphic effectiveness of a ﬂood depends on the duration of ﬂow(s) that exceed the erosional resistance of the channel boundaries, but erosional resistance is a spatially and temporally variable factor that is difﬁcult to quantify except as a mean state or a probability of exceedance. Attempts to quantify the sediment transport component of the geomorphic role of ﬂoods in resistantboundary channels have generally relied on empirical data of bed-material load transport in relation to discharge magnitude, combined with ﬂow frequency and duration data (Andrews and Nankervis, 1995). These types of relations can be successfully applied across hydroclimatically similar regions (e.g., Surian and Andrews, 1999). Attempts to quantify the landform-modiﬁcation component of the geomorphic role of ﬂoods are more likely to use hydraulic modeling in combination with estimated thresholds for processes such as clast entrainment or bedrock quarrying to specify a threshold discharge. Threshold discharge is then used in the context of ﬂow frequency and/or duration to quantify the role of large ﬂoods in modifying speciﬁc ﬂuvial landforms (e.g., Baker, 1977; Wohl, 1992a,b; Baker and Kale, 1998; Kale and Hire, 2004). Whether addressing primarily sediment transport or landform modiﬁcation, quantiﬁcation of the geomorphic role of multiple ﬂoods through time is complicated by the fact that channel conﬁguration (e.g., planform, bedforms, grain-size distribution, gradient) can change substantially during a ﬂood, potentially creating a constantly changing rather than stable relationship between discharge and sediment transport, or between discharge and erosional thresholds. Most assessments of the geomorphic impact of ﬂoods in resistant-boundary channels remain imprecise generalizations. These generalizations are nonetheless useful in recognizing the range of geomorphic and ecological river processes and forms that rely on the occurrence of ﬂoods. As

Review of effects of large floods in resistant-boundary channels

197

more case studies are added to the literature, the scientiﬁc community is getting closer to being able to specify numerical values on the axes in Fig. 8.1, and to being able to quantitatively characterize the geomorphic role of speciﬁc ﬂood magnitudes in a variety of channel settings. Generalized, basin-wide trends in the geomorphic importance of ﬂoods can be drawn from the literature reviewed in this paper, as illustrated in Fig. 8.3. This ﬁgure subdivides an idealized mountainous drainage basin into three general categories. The headwater channels are most likely to have small drainage areas, close connections to adjacent hillslopes, minimal development of ﬂoodplains, steep gradients, and very high boundary resistance associated with very coarse clasts, bedrock, or in-channel wood. Local exceptions to these generalizations certainly occur; mountainous channel networks are commonly longitudinally segmented, for example, creating downstream alternations between lower and higher gradient reaches (Wohl, 2000b). However, the general characteristics listed above lead to frequent disturbances from events such as debris ﬂows and ﬂoods. These headwater channels correspond to the cascade, steppool, and plane-bed channel types of Montgomery and Bufﬁngton (1997), which are supply-limited reaches that are resilient to changes in discharge and sediment supply, and are thus designated transport reaches. In the middle section of Fig. 8.3a, the lower gradient channels have progressively larger drainage areas, greater development of ﬂoodplains, lower gradients, and high boundary resistance associated with moderate-sized clasts (cobble, pebble) and riparian vegetation. These channels correspond to Montgomery and Bufﬁngton’s (1997) planebed and pool-rifﬂe channel types. These response reaches are transport limited and more likely to have sustained responses to changes in sediment supply and discharge. The downstream-most section of Fig. 8.3a includes low-gradient channels with extensive ﬂoodplain development, very large drainage areas, and ﬁner sediment (pebble and smaller) that produces relatively low boundary resistance. These channels correspond to Montgomery and Bufﬁngton’s (1997) pool-rifﬂe and regime-bed (dune-ripple) channel types, and are also transport-limited response reaches. The presence of extensive ﬂoodplains can moderate channel response to a large ﬂood by providing a greater area for sediment deposition and ﬂow energy dissipation than is present in the middle reaches of a drainage basin. The boundaries between these three sections in a given drainage basin could be deﬁned based on drainage area (Clark et al., 1987), stream order (Froehlich et al., 1990), stream gradient (Shroba et al., 1979), or channel type (Montgomery and Bufﬁngton, 1997). These four characteristics are likely to have substantial overlap (e.g., channels with small drainage areas are more likely to have low stream orders, steep gradients, and cascade or step-pool morphology), thus facilitating the zonation of the geomorphic role of ﬂoods within the drainage basin. Returning to the characteristics of rivers dominated by lower frequency ﬂoods (Kochel, 1988), Fig. 8.3b schematically illustrates the downstream trends in these characteristics within a drainage basin. The ﬁrst two characteristics, large seasonal and interannual ﬂow variability, and a high ratio between the discharges of infrequent ﬂoods and average annual ﬂow, are partly dependent on hydroclimatology. Both of these characteristics are likely, however, to reach a maximum in the intermediate portion of a drainage basin. The smaller drainage areas of channel segments in the

E. Wohl

198 (a)

steepest headwater channels: frequent disturbances from floods and debris flows; very high boundary resistance; large supply of coarse sediment; large ratio Qf/Qb;more likely to be dominated by relatively frequent large floods lower gradient channels: infrequent disturbances from floods and debris flows; high boundary resistance; moderate supply of coarse sediment; large ratio Qf/Qb; more likely to be dominated by infrequent floods lowest gradient channels: infrequent disturbances from floods; low boundary resistance; minimal supply of coarse sediment; small ratio Qf/Qb; likely to be dominated by frequent, smaller floods

(b) Flow characteristics

discharge ratio of infrequent flood to average annual flow

seasonal & interannual flow variability

Elevation

Maximum grain size

Distance downstream

Channel-boundary resistance

Distance downstream

riparian vegetation as a bank stabilizer

bank stability resulting from grain size distribution Distance downstream

Review of effects of large floods in resistant-boundary channels

199

upper basin likely result in a higher percentage of contributing area during any given precipitation-generating event, resulting in less difference between large and normal ﬂoods. Conversely, the large drainage areas of channel segments in the lowest portions of the basin likely result in some portion of the upstream basin providing little or no contribution even during extreme events, and/or being attenuated by ﬂoodplain storage, again reducing the difference between large and normal ﬂoods (e.g., Clark et al., 1987). The third characteristic, coarse bedload, declines steadily downstream as stream gradient and proximity to steep valley side slopes decline. Boundary resistance is more complex in that it depends on bank grain-size distribution (including stratigraphy) and riparian vegetation. Bank grain-size distribution declines steadily downstream, but associated resistance rises as sediments become sufﬁciently ﬁne to exhibit cohesion. The effectiveness of riparian vegetation in stabilizing banks is likely to be very region (i.e., climate) speciﬁc, but can be assumed to increase downstream as a ﬁrst approximation. The net result of these trends is that the aggregate population of large ﬂoods is likely to geomorphically dominate the headwater portions of drainage basins, whereas individual large ﬂoods are likely to be most geomorphically important in the middle portion of drainage basins. This broad generalization will have many sitespeciﬁc exceptions, but case studies quantifying the downstream trends in Fig. 8.3 can help to determine the usefulness of this conceptual model. Resistant-boundary channels are most likely to be present in the upper and middle portions of a basin, which correspond to the regions where unit and total stream power, respectively, are hypothesized to reach a maximum value (Knighton, 1999). As a ﬁrst approximation, the relative geomorphic importance of large ﬂoods in resistant-boundary channels can thus be generalized based on position within the drainage basin. There remains a Figure 8.3. (a) Schematic illustration of the distribution of characteristics that govern the geomorphic impact of ﬂoods across a drainage basin. The dark upper line represents the drainage divide, with a channel network developed downstream. In this idealized basin, the low-order channels in the uppermost basin (dark shading) have high boundary resistance associated with bedrock and very large clasts; a steep, laterally conﬁned valley and channel geometry; and abundant coarse sediment supply. The ratio of ﬂood ﬂow to base ﬂow (Qf/Qb) is large. Because of the small contributing area, ﬂoods are relatively frequent, and debris ﬂows from the valley side slopes also disturb the channels. Proceeding downstream, gravel-bed channels (light shading) replace boulder-bed and bedrock channels. The gravel-bed channels have lower boundary resistance and a less abundant supply of coarse sediment than channels higher in the drainage basin. They also have lower gradients and less lateral conﬁnement. The ratio of ﬂood ﬂow to base ﬂow remains large. As contributing area increases, individual ﬂood-generating storm systems are less likely to cover the entire drainage area, so large ﬂoods become less frequent. Extreme storms are still capable of covering much of the upstream area, however, and infrequent large ﬂoods can produce substantial erosional and depositional changes that subsequent, smaller ﬂows are incapable of modifying. In the downstream-most portions of the drainage basin, low channel gradients, limited supply of coarse sediment, limited lateral conﬁnement, and contributing areas too large to be covered by all but the most extreme and infrequent storms, all contribute to a decreasing incidence and geomorphic role of extreme ﬂoods. Superimposed on these trends are the frequency and duration of ﬂoods, the ratio of extreme ﬂoods to base ﬂows, and the time since the last major ﬂood, as these are inﬂuenced by hydroclimatology. (b) Schematic downstream trends in characteristics associated with channels dominated by large, infrequent ﬂoods. The shape of these curves is purely conceptual and is not meant to resemble power functions. The details of how grain-size distribution and riparian vegetation inﬂuence bank stability, for example, can be complex and site-speciﬁc.

200

E. Wohl

strong need for quantiﬁcation of patterns of ﬂood effects in resistant-boundary channels with respect to hydroclimatology, position within a drainage basin, erosional thresholds, sediment supply, and time since the last ﬂood.

Acknowledgements This paper beneﬁted substantially from reviews by Sara Rathburn, Jim O’Connor, and two anonymous reviewers.

References Abbe, T.B., Montgomery, D.R., 2003. Patterns and processes of wood debris accumulation in the Queets River basin, Washington. Geomorphology 51, 81–107. Allred, T.M., Schmidt, J.C., 1999. Channel narrowing by vertical accretion along the Green River near Green River, Utah. Geol. Soc. Am. Bull. 111, 1757–1772. Andrews, E.D., 1984. Bed-material entrainment and hydraulic geometry of gravel-bed rivers in Colorado. Geol. Soc. Am. Bull. 95, 371–383. Andrews, E.D., Erman, D.C., 1986. Persistence in the size distribution of surﬁcial bed material during an extreme snowmelt ﬂood. Water Resour. Res. 22, 191–197. Andrews, E.D., Nankervis, J.M., 1995. Effective discharge and the design of channel maintenance ﬂows for gravel-bed rivers. In: Costa, J.E., Miller, A.J., Potter, K.W., and Wilcock, P.R. (Eds), Natural and Anthropogenic Inﬂuences in Fluvial Geomorphology. American Geophysical Union Press, Washington, DC, pp. 151–164, Geophysical Monograph 89. Anthony, E.J., Julian, M., 1999. Source-to-sink sediment transfers, environmental engineering and hazard mitigation in the steep Var River catchment, French Riviera, southeastern France. Geomorphology 31, 337–354. Auble, G.T., Friedman, J.M., Scott, M.L., 1994. Relating riparian vegetation to present and future streamﬂows. Ecol. Appl. 4, 544–554. Baillie, B.R., Davies, T.R., 2002. Inﬂuence of large woody debris on channel morphology in native forest and pine plantation streams in the Nelson region, New Zealand. New Zeal. J. Mar. Freshwat. Res. 36, 763–774. Baker, V.R., 1973. Paleohydrology and sedimentology of Lake Missoula ﬂooding in eastern Washington. Geol. Soc. Am. Spec. Pap. 144, 79. Baker, V.R., 1977. Stream-channel response to ﬂoods, with examples from central Texas. Geol. Soc. Am. Bull. 88, 1057–1071. Baker, V.R., 1978. Large-scale erosional and depositional features of the Channeled Scabland. In: Baker, V.R. and Nummedal, D. (Eds), The Channeled Scabland. NASA, Washington, DC, pp. 81–115. Baker, V.R., 1984. Flood sedimentation in bedrock ﬂuvial systems. In: Koster, E.H. and Steel, R.J. (Eds), Sedimentology of Gravels and Conglomerates. Can. Soc. Petr. Geol. Mem., Canada 10, pp. 87–98. Baker, V.R., 1987. Paleoﬂood hydrology and extraordinary ﬂood events. J. Hydrol. 96, 79–99. Baker, V.R., Kale, V.S., 1998. The role of extreme ﬂoods in shaping bedrock channels. In: Tinkler, K.J. and Wohl, E.E. (Eds), Rivers Over Rock: Fluvial Processes in Bedrock Channels. American Geophysical Union Press, Washington, DC, pp. 153–165, Geophysical Monograph 107. Baker, V.R., Kochel, R.C., Patton, P.C. (Eds), 1988. Flood Geomorphology. Wiley, New York. Baker, V.R., Pickup, G., 1987. Flood geomorphology of the Katherine Gorge, Northern Territory, Australia. Geol. Soc. Am. Bull. 98, 635–646. Benda, L., Dunne, T., 1997. Stochastic forcing of sediment supply to channel networks from landsliding and debris ﬂow. Water Resour. Res. 33, 2849–2863.

Review of effects of large floods in resistant-boundary channels

201

Benda, L., Veldhuisen, C., Black, J., 2003. Debris ﬂows as agents of morphological heterogeneity at low-order conﬂuences, Olympic Mountains, Washington. Geol. Soc. Am. Bull. 115, 1110–1121. Benito, G., Grodek, T., Enzel, Y., 1998. The geomorphic and hydrologic impacts of the catastrophic failure of ﬂood-control-dams during the 1996-Biescas ﬂood (central Pyrenees, Spain). Z. Geomorphol. 42, 417–437. Bennett, S.J., Pirim, T., Barkdoll, B.D., 2002. Using simulated emergent vegetation to alter stream ﬂow direction within a straight experimental channel. Geomorphology 44, 115–126. Beven, K., 1981. The effect of ordering on the geomorphic effectiveness of hydrologic events. In: Davies, T.R.H. and Pearce, A.J. (Eds), Erosion and Sediment Transport in Paciﬁc Rim Steeplands. IAHS-AISH Publication No. 132, Wallingford, UK, pp. 510–526. Beven, K., Carling, P. (Eds), 1989. Floods: Hydrological, Sedimentological and Geomorphological Implications. John Wiley and Sons, Chichester, UK. Birkeland, G.H., 1996. Riparian vegetation and sandbar morphology along the lower Little Colorado River, Arizona. Phys. Geogr. 17, 534–553. Booker, D.J., Sear, D.A., Payne, A.J., 2001. Modelling three-dimensional ﬂow structures and patterns of boundary shear stress in a natural pool-rifﬂe sequence. Earth Surf. Process. Landf. 26, 553–576. Booth, D.B., 1990. Stream-channel incision following drainage-basin urbanization. Water Resour. Bull. 26, 407–417. Brookes, A., 1988. Channelized rivers: Perspectives for environmental management. John Wiley and Sons, Chichester, UK. Brooks, A.P., Brierley, G.J., Millar, R.G., 2003. The long-term control of vegetation and woody debris on channel and ﬂood-plain evolution: Insights from a paired catchment study in southeastern Australia. Geomorphology 51, 7–29. Brunke, M., Gonser, T., 1997. The ecological signiﬁcance of exchange processes between rivers and groundwater. Freshwat. Biol. 37, 1–33. Brunsden, D., Thornes, J.B., 1979. Landscape sensitivity and change. Trans. Inst. Br. Geogr. New Ser. 4, 463–484. Bufﬁngton, J.M., Lisle, T.E., Woodsmith, R.D., Hilton, S., 2002. Controls on the size and occurrence of pools in coarse-grained forest rivers. River Res. Appl. 18, 507–531. Bufﬁngton, J.M., Montgomery, D.R., 1999. Effects of hydraulic roughness on surface textures of gravelbed rivers. Water Resour. Res. 35, 3507–3521. Bull, W.B., 1979. The threshold of critical power in streams. Geol. Soc. Am. Bull. 90, 453–464. Bunn, J.T., Montgomery, D.R., 2004. Patterns of wood and sediment storage along debris-ﬂow impacted headwater channels in old-growth and industrial forests of the Western Olympic Mountains, Washington. In: Bennett, S.J. and Simon, A. (Eds), Riparian Vegetation and Fluvial Geomorphology. American Geophysical Union Press, Washington, DC, pp. 99–112. Bunn, S.E., Arthington, A.H., 2002. Basic principles and ecological consequences of altered ﬂow regimes for aquatic biodiversity. Environ. Manage. 30, 492–507. Burbank, D.W., Leland, J., Fielding, E., et al., 1996. Bedrock incision, rock uplift and threshold hillslopes in the northwestern Himalayas. Nature 379, 505–510. Burkham, D.E., 1972. Channel changes of the Gila River in Safford Valley, Arizona, 1846–1970. US Geol. Surv. Prof. Pap. 655G, 24. Carling, P.A., 1987. Hydrodynamic interpretation of a boulder berm and associated debris-torrent deposits. Geomorphology 1, 53–67. Carling, P.A., Glaister, M.S., 1987. Reconstruction of a ﬂood resulting from a moraine-dam failure using geomorphological evidence and dam-break modeling. In: Mayer, L. and Nash, D. (Eds), Catastrophic Flooding. Allen and Unwin, Boston, MA, pp. 182–200. Carling, P.A., Reader, N.A., 1982. Structure, composition and bulk properties of upland stream gravels. Earth Surf. Process. Landf. 7, 349–365. Carling, P.A., Tinkler, K.J., 1998. Conditions for entrainment of cuboid boulders in bedrock streams: An historical review of literature with respect to recent investigations. In: Tinkler, K.J. and Wohl, E.E. (Eds), Rivers Over Rock: Fluvial Processes in Bedrock Channels. American Geophysical Union Press, Washington, DC, pp. 19–34, Geophysical Monograph 107. Carson, M.A., Grifﬁths, G.A., 1987. Bedload transport in gravel channels. J. Hydrol. (NZ) 26, 1–151.

202

E. Wohl

Cenderelli, D.A., 2000. Floods from natural and artiﬁcial dam failures. In: Wohl, E.E. (Ed.), Inland Flood Hazards: Human, Riparian, and Aquatic Communities. Cambridge University Press, Cambridge, UK, pp. 73–103. Cenderelli, D.A., Cluer, B.L., 1998. Depositional processes and sediment supply in resistant-boundary channels: Examples from two case studies. In: Tinkler, K.J. and Wohl, E.E. (Eds), Rivers Over Rock: Fluvial Processes in Bedrock Channels. American Geophysical Union Press, Washington, DC, pp. 105–131, Geophysical Monograph 107. Cenderelli, D.A., Wohl, E.E., 1998. Sedimentology and clast orientation of deposits produced by glacial-lake outburst ﬂoods in the Mount Everest region, Nepal. In: Kalvoda, J. and Rosenfeld, C.L. (Eds), Geomorphological Hazards in High Mountain Areas. Kluwer Academic Publishers, Dordrecht, pp. 1–26. Cenderelli, D.A., Wohl, E.E., 2001. Peak discharge estimates of glacial-lake outburst ﬂoods and ‘‘normal’’ climatic ﬂoods in the Mount Everest region, Nepal. Geomorphology 40, 57–90. Cenderelli, D.A., Wohl, E.E., 2003. Flow hydraulics and geomorphic effects of glacial-lake outburst ﬂoods in the Mount Everest region, Nepal. Earth Surf. Process. Landf. 28, 385–407. Chin, A., Gregory, K.J., 2001. Urbanization and adjustment of ephemeral stream channels. Ann. Assoc. Am. Geogr. 91, 595–608. Church, M., 1988. Floods in cold climates. In: Baker, V.R., Kochel, R.C., and Patton, P.C. (Eds), Flood Geomorphology. John Wiley and Sons, New York, pp. 205–229. Ciszewski, D., 2001. Flood-related changes in heavy metal concentrations within sediments of the Bia"a Przemsza River. Geomorphology 40, 205–218. Clark, G.M., Jacobson, R.B., Kite, J.S., Linton, R.C., 1987. Storm-induced catastrophic ﬂooding in Virginia and West Virginia, November, 1985. In: Mayer, L. and Nash, D. (Eds), Catastrophic Flooding. Allen and Unwin, Boston, MA, pp. 355–379. Clark, J.J., Wilcock, P.R., 2000. Effects of land-use change on channel morphology in northeastern Puerto Rico. Geol. Soc. Am. Bull. 112, 1763–1777. Collier, M., Webb, R.H., Schmidt, J.C., 1996. Dams and rivers: Primer on the downstream effects of dams. US Geol. Surv. Circular 1126, 94. Collins, B.D. and Montgomery, D.R., 2001. Importance of archival and process studies to characterizing pre-settlement riverine geomorphic processes and habitat in the Puget Lowland. In: Dorava, J.M., Montgomery, D.R., Palcsak, B.B., and Fitzpatrick, F.A. (Eds), Geomorphic Processes and Riverine Habitat. American Geophysical Union Press, Washington, DC, pp. 227–243. Costa, J.E., 1987. Hydraulics and basin morphometry of the largest ﬂash ﬂoods in the conterminous United States. J. Hydrol. 93, 313–338. Costa, J.E., O’Connor, J.E., 1995. Geomorphically effective ﬂoods. In: Costa, J.E., Miller, A.J., Potter, K.W., and Wilcock, P.R. (Eds), Natural and Anthropogenic Inﬂuences in Fluvial Geomorphology. American Geophysical Union Press, Washington, DC, pp. 45–56, Geophysical Monograph 89. Costa, J.E., Schuster, R.L., 1988. The formation and failure of natural dams. Geol. Soc. Am. Bull. 100, 1054–1068. Cui, Y., Parker, G., Lisle, T.E., et al., 2003a. Sediment pulses in mountain rivers: Part 1. Experiments. Water Resour. Res. 39, 1239. Cui, Y., Parker, G., Pizzuto, J., Lisle, T.E., 2003b. Sediment pulses in mountain rivers: Part 2. Comparison between experiments and numerical predictions. Water Resour. Res. 39, 1240. Curran, J.H., Wohl, E.E., 2003. Large woody debris and ﬂow resistance in step-pool channels, Cascade Range, Washington. Geomorphology 51, 141–157. De Jong, C., 1994. The signiﬁcance of extreme events in the development of mountain river beds. In: Olive, L.J., Loughran, R.J. and Kesby, J.A. (Eds), Variability in Stream Erosion and Sediment Transport. IAHS Publication No. 224, Wallingford, UK, pp. 13–24. De Jong, C., Ergenzinger, P., 1995. The interrelations between mountain valley form and river-bed arrangement. In: Hickin, E.J. (Ed.), River Geomorphology. John Wiley and Sons, Chichester, UK, pp. 55–91. Desloges, J.R., Church, M., 1992. Geomorphic implications of glacier outburst ﬂooding: Noeick River valley, British Columbia. Can. J. Earth Sci. 29, 551–564. Dietrich, W.E., Kirchner, J.W., Ikeda, H., Iseya, F., 1989. Sediment supply and the development of the coarse surface layer in gravel-bedded rivers. Nature 340, 215–217.

Review of effects of large floods in resistant-boundary channels

203

Dubinski, I.M., 2005. Historical sediment budget for the Black Canyon of the Gunnison National Park, Colorado. Unpublished MS Thesis, Colorado State University, Fort Collins, CO, 201pp. Eaton, B.C., Lapointe, M.F., 2001. Effects of large ﬂoods on sediment transport and reach morphology in the cobble-bed Sainte Marguerite River. Geomorphology 40, 291–309. Elliott, J.G., Hammack, L.A., 2000. Entrainment of riparian gravel and cobbles in an alluvial reach of a regulated canyon river. Regul. Rivers Res. Manage. 16, 37–50. Elliott, J.G., Parker, R.S., 1997. Altered streamﬂow and sediment entrainment in the Gunnison Gorge. Water Resour. Bull. 33, 1041–1054. Erman, D.C., Andrews, E.D., Yoder-Williams, M., 1988. Effects of winter ﬂoods on ﬁshes in the Sierra Nevada. Can. J. Fish. Aquat. Sci. 45, 2195–2200. Eschner, T.R., Hadley, R.F., Crowley, K.D., 1983. Hydrologic and morphologic changes in channels of the Platte River basin in Colorado, Wyoming, and Nebraska: A historical perspective. US Geol. Surv. Prof. Pap. 1277-A, A1–A39. Everitt, B.L., 1968. Use of the cottonwood in an investigation of the recent history of a ﬂood plain. Am. J. Sci. 266, 417–439. Fausch, K.D., Bramblett, R.G., 1991. Disturbance and ﬁsh communities in intermittent tributaries of a western Great Plains river. Copeia 1991 (3), 659–674. Fausch, K.D., Northcote, T.G., 1992. Large woody debris and salmonid habitat in a small coastal British Columbia stream. Can. J. Fish. Aquat. Sci. 49, 682–693. Fausch, K.D., Torgersen, C.E., Baxter, C.V., Li, H.W., 2002. Landscapes to riverscapes: Bridging the gap between research and conservation of stream ﬁshes. BioScience 52, 483–498. Faustini, J.M., Jones, J.A., 2003. Inﬂuence of large woody debris on channel morphology and dynamics in steep, boulder-rich mountain streams, western Cascades, Oregon. Geomorphology 51, 187–205. Fisher, S.G., Grimm, N.B., Marti, E., Gomez, R., 1998. Hierarchy, spatial conﬁguration, and nutrient cycling in a desert stream. Aus. J. Ecol. 23, 41–52. Fonstad, M., Marcus, W.A., 2003. Self-organized criticality in riverbank systems. Ann. Assoc. Am. Geogr. 93, 281–296. Friedman, J.M., Osterkamp, W.R., Lewis, W.M., 1996. The role of vegetation and bed-level ﬂuctuations in the process of channel narrowing. Geomorphology 14, 341–351. Froehlich, W., 1991. Sediment production from unmetalled road surfaces. In: Peters, N.E. and Walling, D.E. (Eds), Sediment and Stream Water Quality in a Changing Environment: Trends and Explanation. IAHS Publication No. 203, Wallingford, UK, pp. 21–29. Froehlich, W., Gil, E., Kasza, I., Starkel, L., 1990. Thresholds in the transformation of slopes and river channels in the Darjeeling Himalaya, India. Mt. Res. Dev. 10, 301–312. Froehlich, W., Starkel, L., 1987. Normal and extreme monsoon rains – their role in the shaping of the Darjeeling Himalaya. Stud. Geomorphol. Carpatho-Balcanica 21, 129–158. Gavrilovic, Z., Matovic, Z., 1991. Review of disastrous torrent ﬂood on the Vlasina River on June 26, 1988 – including analysis of ﬂood and the obtained results. In: Armanini, A. and DiSilvio, G. (Eds), Fluvial Hydraulics of Mountain Regions. Springer-Verlag, Berlin, pp. 235–250. Gomez, B., 1983. Temporal variations in bedload transport rates: The effect of progressive bed armouring. Earth Surf. Process. Landf. 8, 41–54. Gomez, B., Church, M., 1989. An assessment of bed load sediment transport formulae for gravel bed rivers. Water Resour. Res. 25, 1161–1186. Gomez, B., Russell, A.J., Smith, L.C. and Knudsen, O., 2002. Erosion and deposition in the proglacial zone: the 1996 jo¨kulhlaup on Skeioararsandur, southeast Iceland. In: Snorrason, A., Finnsdottir, H.P. and Moss, M. (Eds), The Extremes of the Extremes: Extraordinary Floods. IAHS Publication No. 271, Wallingford, UK, pp. 217–221. Gomi, T., Sidle, R.C., Woodsmith, R.D., Bryant, M.D., 2003. Characteristics of channel steps and reach morphology in headwater streams, southeast Alaska. Geomorphology 51, 225–242. Graf, W.L., 1978. Fluvial adjustments to the spread of tamarisk in the Colorado Plateau region. Geol. Soc. Am. Bull. 89, 1491–1501. Graf, W.L., 1988. Fluvial processes in dryland rivers. Springer-Verlag, Berlin. Graf, W.L., 1999. Dam nation: A geographic census of American dams and their large-scale hydrologic impacts. Water Resour. Res. 35, 1305–1311.

204

E. Wohl

Graf, W.L., 2001. Damage control: Restoring the physical integrity of America’s rivers. Ann. Assoc. Am. Geogr. 91 (1), 1–27. Grant, G.E., Schmidt, J.C., Lewis, S.L., 2003. A geological framework for interpreting downstream effects of dams on rivers. In: O’Connor, J.E. and Grant, G.E. (Eds), A Peculiar River: Geology, Geomorphology, and Hydrology of the Deschutes River, Oregon. American Geophysical Union Press, Washington, DC, pp. 203–219. Grant, G.E., Swanson, F.J., Wolman, M.G., 1990. Pattern and origin of stepped-bed morphology in highgradient streams, western Cascades, Oregon. Geol. Soc. Am. Bull. 102, 340–352. Gray, D.H., Barker, D., 2004. Root-soil mechanics and interactions. In: Bennett, S.J. and Simon, A. (Eds), Riparian Vegetation and Fluvial Geomorphology. American Geophysical Union Press, Washington, DC, pp. 113–123. Gupta, A., 1983. High-magnitude ﬂoods and stream channel response. Spec. Publs. Int. Assoc. Sedimentol. 6, 219–227. Gupta, A., 1988. Large ﬂoods as geomorphic events in humid tropics. In: Baker, V.R., Kochel, R.C., and Patton, P.C. (Eds), Flood Geomorphology. John Wiley and Sons, New York, pp. 301–320. Gupta, A., Dutt, A., 1989. The Auranga: Description of a tropical monsoon river. Z. Geomorphol. 33, 73–92. Hancock, G.S., Anderson, R.S., Whipple, K.X., 1998. Beyond power: Bedrock river incision process and form. In: Tinkler, K.J. and Wohl, E.E. (Eds), Rivers Over Rock: Fluvial Processes in Bedrock Channels. American Geophysical Union Press, Washington, DC, pp. 35–60, Geophysical Monograph 107. Harvey, A.M., 1984. Geomorphological response to an extreme ﬂood: A case from southeast Spain. Earth Surf. Process. Landf. 9, 267–279. Harvey, M.D., Mussetter, R.A., Wick, E.J., 1993. A physical process – biological response model for spawning habitat formation for the endangered Colorado squawﬁsh. Rivers 4, 114–131. Heede, B.H., 1985. Channel adjustments to the removal of log steps: An experiment in a mountain stream. Environ. Manage. 9, 427–432. Hewitt, K., 1998. Catastrophic landslides and their effects on the Upper Indus streams, Karakoram Himalaya, northern Pakistan. Geomorphology 26, 47–80. Hilmes, M.M., Wohl, E.E., 1995. Changes in channel morphology associated with placer mining. Phys. Geogr. 16, 223–242. Hirschboeck, K.K., 1987. Catastrophic ﬂooding and atmospheric circulation anomalies. In: Mayer, L. and Nash, D. (Eds), Catastrophic Flooding. Allen and Unwin, Boston, MA, pp. 25–56. Hirschboeck, K.K., 1988. Flood hydroclimatology. In: Baker, V.R., Kochel, R.C., and Patton, P.C. (Eds), Flood Geomorphology. John Wiley and Sons, New York, pp. 27–49. Howard, A.D., 1998. Long proﬁle development of bedrock channels: Interaction of weathering, mass wasting, bed erosion, and sediment transport. In: Tinkler, K.J. and Wohl, E.E. (Eds), Rivers Over Rock: Fluvial Processes in Bedrock Channels. American Geophysical Union Press, Washington, DC, pp. 297–319, Geophysical Monograph 107. Hupp, C.R., Osterkamp, W.R., 1996. Riparian vegetation and ﬂuvial geomorphic process. Geomorphology 14, 277–295. Inbar, M., 1987. Effects of a high magnitude ﬂood in a Mediterranean climate: A case study in the Jordan River basin. In: Mayer, L. and Nash, D. (Eds), Catastrophic Flooding. Allen and Unwin, Boston, MA, pp. 333–353. Jacobson, R.B., Miller, A.J., Smith, J.A., 1989. The role of catastrophic geomorphic events in central Appalachian landscape evolution. Geomorphology 2, 257–284. James, A., 1999. Time and the persistence of alluvium: River engineering, ﬂuvial geomorphology, and mining sediment in California. Geomorphology 31, 265–290. Jansen, J.D., Brierley, G.J., 2004. Pool-ﬁlls: A window to palaeoﬂood history and response in bedrockconﬁned rivers. Sedimentology 51, 901–925. Jaquette, C., Wohl, E., Cooper, D., 2005. Establishing a context for river rehabilitation, North Fork Gunnison River, Colorado. Environ. Manage. 35, 593–606. Jeffries, R., Darby, S.E., Sear, D.A., 2003. The inﬂuence of vegetation and organic debris on ﬂood-plain sediment dynamics: Case study of a low-order stream in the New Forest, England. Geomorphology 51, 61–80.

Review of effects of large floods in resistant-boundary channels

205

Johansson, M.E., Nilsson, C., 2002. Responses of riparian plants to ﬂooding in free-ﬂowing and regulated boreal rivers: An experimental study. J. Appl. Ecol. 39, 971–986. Johnson, W.C., 1994. Woodland expansion in the Platte River, Nebraska: Patterns and causes. Ecol. Monogr. 64, 45–84. Junk, W.J., Bayley, P.B. and Sparks, R.E., 1989. The ﬂood pulse concept in river-ﬂoodplain systems. In: Dodge, D.P. (Ed.), Proceedings of the International Large River Symposium. Can. Spec. Publ. Fish. Aquat. Sci. 106, 110–127. Kail, J., 2003. Inﬂuence of large woody debris on the morphology of six central European streams. Geomorphology 51, 207–223. Kale, V.S., Hire, P.S., 2004. Effectiveness of monsoon ﬂoods on the Tapi River, India: Role of channel geometry and hydrologic regime. Geomorphology 57, 275–291. Kasahara, T., Wondzell, S.M., 2003. Geomorphic controls on hyporheic exchange ﬂow in mountain streams. Water Resour. Res. 39, SBH 3-1 to 3-14. Kean, J.W., Smith, J.D., 2004. Flow and boundary shear stress in channels with woody bank vegetation. In: Bennett, S.J. and Simon, A. (Eds), Riparian Vegetation and Fluvial Geomorphology. American Geophysical Union Press, Washington, DC, pp. 237–252. Keller, E.A., Swanson, F.J., 1979. Effects of large organic material on channel form and ﬂuvial processes. Earth Surf. Process. Landf. 4, 361–380. Keller, E.A., Tally, T., 1979. Effects of large organic debris on channel form and ﬂuvial processes in the coastal redwood environment. In: Rhodes, D.D. and Williams, G.P. (Eds), Adjustments of the Fluvial System. Kendall/Hunt Publishing, Dubuque, IA, pp. 169–197. Kieffer, S.W., 1985. The 1983 hydraulic jump in Crystal Rapid: Implications for river-running and geomorphic evolution in the Grand Canyon. J. Geol. 93, 385–406. Kirchner, J.W., Dietrich, W.E., Iseya, F., Ikeda, H., 1990. The variability of critical shear stress, friction angle, and grain protrusion in water-worked sediments. Sedimentology 37, 647–672. Kite, J.S. and Linton, R.C., 1993. Depositional aspects of the November 1985 ﬂood on Cheat River and Black Fork, West Virginia. In: Jacobson, R.B. (Ed.), Geomorphic Studies of the Storm and Flood of November 3–5, 1985, in the Upper Potomac and Cheat River Basins in West Virginia and Virginia. US Geol. Surv. Bull. 1981, D1–D24. Klein, R., Sonnevil, R., Short, D., 1987. Effects of woody debris removal on sediment storage in a northwest California stream. In: Beschta, R.L., Blinn, T., Grant, G.E., Ice, G.G. and Swanson, F.J. (Eds), Erosion and Sedimentation in the Paciﬁc Rim, IAHS Publication No. 165, Wallingford, UK, pp. 403–404. Klimek, K., 1987. Man’s impact on ﬂuvial processes in the Polish Western Carpathians. Geogr. Ann. 69A, 221–229. Knighton, A.D., 1989. River adjustment to changes in sediment load: The effects of tin mining on the Ringarooma River, Tasmania, 1875–1984. Earth Surf. Process. Landf. 14, 333–359. Knighton, A.D., 1999. Downstream variation in stream power. Geomorphology 29, 293–306. Kochel, R.C., 1988. Geomorphic impact of large ﬂoods: Review and new perspectives on magnitude and frequency. In: Baker, V.R., Kochel, R.C., and Patton, P.C. (Eds), Flood Geomorphology. John Wiley and Sons, New York, pp. 169–187. Kodama, Y., 1994. Downstream changes in the lithology and grain size of ﬂuvial gravels, the Watarase River, Japan: Evidence of the role of abrasion in downstream ﬁning. J. Sediment. Res. A 64, 68–75. Kondolf, G.M., 1996. A cross section of stream channel restoration. J. Soil Water Conserv. 51, 119–125. Kondolf, G.M., 1997. Hungry water: Effects of dams and gravel mining on river channels. Environ. Manage. 21, 533–551. Kondolf, G.M., Smeltzer, M.W., Railsback, S., 2001. Design and performance of a channel reconstruction project in a coastal California gravel-bed stream. Environ. Manage. 28, 761–776. Korup, O., 2005. Large landslides and their effect on sediment ﬂux in South Westland, New Zealand. Earth Surf. Process. Landf. 30, 305–323. Kresan, P.L., 1988. The Tucson, Arizona, ﬂood of October 1983. In: Baker, V.R., Kochel, R.C., and Patton, P.C. (Eds), Flood Geomorphology. John Wiley and Sons, New York, pp. 465–489. Lancaster, S.T., Hayes, S.K., Grant, G.E., 2001. Modeling sediment and wood storage and dynamics in small mountainous watersheds. In: Dorava, J.M., Montgomery, D.R., Palcsak, B.B., and Fitzpatrick,

206

E. Wohl

F.A. (Eds), Geomorphic Processes and Riverine Habitat. American Geophysical Union Press, Washington, DC, pp. 85–102. Lapointe, M.F., Secretan, Y., Driscoll, S.N., et al., 1998. Response of the Ha! Ha! River to the ﬂood of July 1996 in the Saguenay Region of Quebec: Large-scale avulsion in a glaciated valley. Water Resour. Res. 34, 2383–2392. Larsen, M.C., Roman, A.S., 2001. Mass wasting and sediment storage in a small montane watershed: An extreme case of anthropogenic disturbance in the humid tropics. In: Dorava, J.M., Montgomery, D.R., Palcsak, B.B., and Fitzpatrick, F.A. (Eds), Geomorphic Processes and Riverine Habitat. American Geophysical Union Press, Washington, DC, pp. 119–138. Lawler, D.M., 1992. Process dominance in bank erosion systems. In: Carling, P.A. and Petts, G.E. (Eds), Lowland Floodplain Rivers: Geomorphological Perspectives. John Wiley and Sons, Chichester, UK, pp. 117–143. Lawless, M., Robert, A., 2001. Three-dimensional ﬂow structure around small-scale bedforms in a simulated gravel-bed environment. Earth Surf. Process. Landf. 26, 507–522. Lenzi, M., 2001. Step-pool evolution in the Rio Cordon, northeastern Italy. Earth Surf. Process. Landf. 26, 991–1008. Lenzi, M.A., Mao, L., Comiti, F., 2004. Magnitude-frequency analysis of bed load data in an Alpine boulder bed stream. Water Resour. Res. 40, WR002961. Leopold, L.B., Wolman, M.G., Miller, J.P., 1964. Fluvial processes in geomorphology. W.H. Freeman and Company, San Francisco, CA. Ligon, F.K., Dietrich, W.E., Trush, W.J., 1995. Downstream ecological effects of dams. BioScience 45, 183–192. Liquori, M.K., 1995. Coarse clast delivery and transport processes in the Black Canyon of the Gunnison River, Colorado. Unpublished MS Thesis, Colorado State University, Fort Collins, CO, 214pp. Lisle, T.E., 1982. Effects of aggradation and degradation on rifﬂe-pool morphology in natural gravel channels, northwestern California. Water Resour. Res. 18, 1643–1651. Lisle, T.E., Pizzuto, J.E., Ikeda, H., Iseya, F., Kodama, Y., 1997. Evolution of a sediment wave in an experimental channel. Water Resour. Res. 33, 1971–1981. MacFarlane, W.A., Wohl, E.E., 2003. Inﬂuence of step composition on step geometry and ﬂow resistance in step-pool streams of the Washington Cascades. Water Resour. Res. 39, WR001238. Macklin, M.G., Rumsby, B.T., Newson, M.D., 1992. Historical ﬂoods and vertical accretion of ﬁnegrained alluvium in the Lower Tyne valley, northeast England. In: Billi, P., Hey, R.D., C.R. Thorne, C.R., and Tacconi, P. (Eds), Dynamics of Gravel-Bed Rivers. John Wiley and Sons, Chichester, UK, pp. 573–589. Madej, M.A., 2001. Development of channel organization and roughness following sediment pulses in single-thread, gravel bed rivers. Water Resour. Res. 37, 2259–2272. Madej, M.A., Ozaki, V., 1996. Channel response to sediment wave propagation and movement, Redwood Creek, California, USA. Earth Surf. Process. Landf. 21, 911–927. Magilligan, F.J., 1992. Thresholds and the spatial variability of ﬂood power during extreme ﬂoods. Geomorphology 5, 373–390. Magilligan, F.J., Nislow, K.H., Graber, B.E., 2003. Scale-independent assessment of discharge reduction and riparian disconnectivity following ﬂow regulation by dams. Geology 31, 569–572. Maizels, J., 1991. The origin and evolution of Holocene sandur deposits in areas of jo¨kulhlaup drainage, Iceland. In: Maizels, J.K. and Caseldine, C. (Eds), Environmental Change in Iceland: Past and Present. Kluwer Academic Publishers, Dordrecht, pp. 267–302. Maizels, J., 1997. Jo¨kulhlaup deposits in proglacial areas. Quaternary Sci. Rev. 16, 793–819. Malde, H.E., 1968. The catastrophic late Pleistocene Bonneville Flood in the Snake River plain, Idaho. US Geol. Surv. Prof. Pap. 596, 52. McKenney, R., 2001. Channel changes and habitat diversity in a warm-water, gravel-bed stream. In: Dorava, J.M., Montgomery, D.R., Palcsak, B.B., and Fitzpatrick, F.A. (Eds), Geomorphic Processes and Riverine Habitat. American Geophysical Union Press, Washington, DC, pp. 57–71. Mei-e, R., Xianmo, Z., 1994. Anthropogenic inﬂuences on changes in the sediment load of the Yellow River, China, during the Holocene. Holocene 4, 314–320.

Review of effects of large floods in resistant-boundary channels

207

Merritt, D.M., Cooper, D.J., 2000. Riparian vegetation and channel change in response to river regulation: A comparative study of regulated and unregulated streams in the Green River basin, USA. Regul. Rivers Res. Manage. 16, 543–564. Merritt, D.M., Wohl, E.E., 2002. Processes governing hydrochory along rivers: Hydraulics, hydrology, and dispersal phenology. Ecol. Appl. 12, 1071–1087. Merritt, D.M., Wohl, E.E., 2003. Downstream hydraulic geometry and channel adjustment during a ﬂood along an ephemeral, arid-region drainage. Geomorphology 52, 165–180. Miller, A.J., 1990. Flood hydrology and geomorphic effectiveness in the central Appalachians. Earth Surf. Process. Landf. 15, 119–134. Miller, A.J., Cluer, B.L., 1998. Modeling considerations for simulation of ﬂow in bedrock channels. In: Tinkler, K.J. and Wohl, E.E. (Eds), Rivers Over Rock: Fluvial Processes in Bedrock Channels. American Geophysical Union Press, Washington, DC, pp. 61–104, Geophysical Monograph 107. Miller, A.J. and Parkinson, D.J., 1993. Flood hydrology and geomorphic effects on river channels and ﬂood plains: the ﬂood of November 4–5, 1985, in the South Branch Potomac River basin of West Virginia. In: Jacobson, R.B. (Ed.), Geomorphic Studies of the Storm and Flood of November 3–5, 1985, in the Upper Potomac and Cheat River Basins in West Virginia and Virginia. US Geol. Surv. Bull. 1981, E1–E96. Miller, J.R., 1991. The inﬂuence of bedrock geology on knickpoint development and channel-bed degradation along downcutting streams in south-central Indiana. J. Geol. 99, 591–605. Montgomery, D.R., Bufﬁngton, J.M., 1997. Channel-reach morphology in mountain drainage basins. Geol. Soc. Am. Bull. 109, 596–611. Montgomery, D.R., Bufﬁngton, J.M., Smith, R.D., Schmidt, K.M., Pess, G., 1995. Pool spacing in forest channels. Water Resour. Res. 31, 1097–1105. Montgomery, D.R., Massong, T.M., Hawley, S.C.S., 2003. Inﬂuence of debris ﬂows and logjams on the location of pools and alluvial channel reaches, Oregon Coast Range. Geol. Soc. Am. Bull. 115, 78–88. Montgomery, J.R., Abbe, T.B., Bufﬁngton, J.M., et al., 1996. Distribution of bedrock and alluvial channels in forested mountain drainage basins. Nature 381, 587–589. Mosley, M.P. (Ed.), 2001. Gravel-Bed Rivers V. International Gravel-Bed Rivers Workshop, Christchurch, New Zealand, August 2000. Nadler, C.T., Schumm, S.A., 1981. Metamorphosis of South Platte and Arkansas Rivers, eastern Colorado. Phys. Geogr. 2, 95–115. Nakamura, F., Shin, N., 2001. The downstream effects of dams on the regeneration of riparian tree species in northern Japan. In: Dorava, J.M., Montgomery, D.R., Palcsak, B.B., and Fitzpatrick, F.A. (Eds), Geomorphic Processes and Riverine Habitat. American Geophysical Union Press, Washington, DC, pp. 173–181. Nelson, J.M., Shreve, R.L., McLean, S.R., Drake, T.G., 1995. Role of near-bed turbulence structure in bed load transport and bed form mechanics. Water Resour. Res. 31, 2071–2086. Nicholas, A.P., 2001. Computational ﬂuid dynamics modeling of boundary roughness in gravel-bed rivers: An investigation of the effects of random variability in bed elevation. Earth Surf. Process. Landf. 26, 345–362. Nilsson, C., Gardfjell, M., Grelsson, G., 1991. Importance of hydrochory in structuring plant communities along rivers. Can. J. Bot. 69, 2631–2633. Nilsson, C., Jansson, R., 1995. Floristic differences between riparian corridors of regulated and free-ﬂowing boreal rivers. Regul. Rivers Res. Manage. 11, 55–66. Nolan, K.M., Marron, D.C., 1985. Contrast in stream-channel response to major storms in two mountainous areas of California. Geology 13, 135–138. Nott, J., Price, D., 1994. Plunge pools and paleoprecipitation. Geology 22, 1047–1050. O’Connor, J.E., 1993. Hydrology, hydraulics, and geomorphology of the Bonneville ﬂood. Geol. Soc. Am. Spec. Pap. 274, Boulder, CO, 83pp. O’Connor, J.E., Costa, J.E., 2004. Spatial distribution of the largest rainfall-runoff ﬂoods from basins between 2.6 and 26,000 km2 in the United States and Puerto Rico. Water Resour. Res. 40, W01107. O’Connor, J.E., Grant, G.E., Costa, J.E., 2002. The geology and geography of ﬂoods. In: House, P.K., Webb, R.H., Baker, V.R., and Levish, D.R. (Eds), Ancient Floods, Modern Hazards: Principles and

208

E. Wohl

Applications of Paleoﬂood Hydrology. American Geophysical Union Press, Washington, DC, pp. 359–385. O’Connor, J.E., Hardison, J.H., Costa, J.E., 2001. Debris ﬂows from failures of Neoglacial-age moraine dams in the Three Sisters and Mount Jefferson Wilderness Areas, Oregon. US Geol. Surv. Prof. Pap. 1606, 93. O’Connor, J.E., Jones, M.A., Haluska, T.L., 2003. Flood plain and channel dynamics of the Quinault and Queets Rivers, Washington, USA. Geomorphology 51, 31–59. O’Connor, J.E., Webb, R.H., Baker, V.R., 1986. Paleohydrology of pool-and-rifﬂe pattern development: Boulder Creek, Utah. Geol. Soc. Am. Bull. 97, 410–420. Parker, G., Klingeman, P.C., 1982. On why gravel bed streams are paved. Water Resour. Res. 18, 1409–1423. Parker, G., Sutherland, A.J., 1990. Fluvial armour. J. Hydraul. Res. 28, 529–544. Parker, G., Toro-Escobar, C.M., 2002. Equal mobility of gravel in streams: The remains of the day. Water Resour. Res. 38, WR000669. Patton, P.C., 1988. Geomorphic response of streams to ﬂoods in the glaciated terrain of southern New England. In: Baker, V.R., Kochel, R.C., and Patton, P.C. (Eds), Flood Geomorphology. John Wiley and Sons, New York, pp. 261–277. Patton, P.C., Baker, V.R., 1976. Morphometry and ﬂoods in small drainage basins subject to diverse hydrogeomorphic controls. Water Resour. Res. 12, 941–952. Pickett, S.T.A., White, P.S. (Eds), 1985. The Ecology of Natural Disturbance and Patch Dynamics. Academic Press, Orlando, FL. Pie´gay, H., Bravard, J.-P., 1997. Response of a Mediterranean riparian forest to a 1 in 400 year ﬂood, Ouveze River, Drome-Vaucluse, France. Earth Surf. Process. Landf. 22, 31–43. Pie´gay, H., Gurnell, A.M., 1997. Large woody debris and river geomorphological pattern: Examples from S.E. France and S. England. Geomorphology 19, 99–116. Pie´gay, H., Salvador, P.-G., 1997. Contemporary ﬂoodplain forest evolution along the middle Ubaye River, Southern Alps, France. Global Ecol. Biogeogr. Lett. 6, 397–406. Pie´gay, H., The´venet, A., Citterio, A., 1999. Input, storage and distribution of large woody debris along a mountain river continuum, the Droˆme River, France. Catena 35, 19–39. Pitlick, J., 1994. Relation between peak ﬂows, precipitation, and physiography for ﬁve mountainous regions in the western USA. J. Hydrol. 158, 219–240. Pitlick, J., Wilcock, P., 2001. Relations between streamﬂow, sediment transport, and aquatic habitat in regulated rivers. In: Dorava, J.M., Montgomery, D.R., Palcsak, B.B., and Fitzpatrick, F.A. (Eds), Geomorphic Processes and Riverine Habitat. American Geophysical Union Press, Washington, DC, pp. 185–198. Poff, N.L., Allan, J.D., Bain, M.B., et al., 1997. The natural ﬂow regime. BioScience 47, 769–784. Pollen, N., Simon, A., 2005. Estimating the mechanical effects of riparian vegetation on stream bank stability using a ﬁber bundle model. Water Resour. Res. 41, W07025. Pollen, N., Simon, A., Collison, A., 2004. Advances in assessing the mechanical and hydrologic effects of riparian vegetation on streambank stability. In: Bennett, S.J. and Simon, A. (Eds), Riparian Vegetation and Fluvial Geomorphology. American Geophysical Union Press, Washington, DC, pp. 125–139. Powell, D.M., 1998. Patterns and processes of sediment sorting in gravel-bed rivers. Progr. Phys. Geogr. 22, 1–32. Power, M.E., Parker, G., Dietrich, W.E., Sun, A., 1995. How does ﬂoodplain width affect ﬂoodplain river ecology? A preliminary exploration using simulations. Geomorphology 13, 301–317. Robert, A., Roy, A.G., De Serres, B., 1996. Turbulence at a roughness transition in a depth limited ﬂow over a gravel bed. Geomorphology 16, 175–187. Roberts, C.R., 1989. Flood frequency and urban-induced channel change: Some British examples. In: Beven, K. and Carling, P. (Eds), Floods: Hydrological, Sedimentological and Geomorphological Implications. John Wiley and Sons, Chichester, UK, pp. 57–82. Rubin, D.M., Nelson, J.M., Topping, D.J., 1998. Relation of inversely graded deposits to suspendedsediment grain-size evolution during the 1996 ﬂood experiment in Grand Canyon. Geology 26, 99–102. Rutherfurd, I.D., Grove, J.R., 2004. The inﬂuence of trees on stream bank erosion: Evidence from rootplate abutments. In: Bennett, S.J. and Simon, A. (Eds), Riparian Vegetation and Fluvial Geomorphology. American Geophysical Union Press, Washington, DC, pp. 141–152.

Review of effects of large floods in resistant-boundary channels

209

Sawada, T., Ashida, K., Takahashi, T., 1983. Relationship between channel pattern and sediment transport in a steep gravel bed river. Z. Geomorphol. 46, 55–66. Schick, A.P., 1988. Hydrologic aspects of ﬂoods in extreme arid environments. In: Baker, V.R., Kochel, R.C., and Patton, P.C. (Eds), Flood Geomorphology. John Wiley and Sons, New York, pp. 189–203. Schmidt, J.C., 1990. Recirculating ﬂow and sedimentation in the Colorado River in Grand Canyon, Arizona. J. Geol. 98, 709–724. Schmidt, J.C., Rubin, D.M., 1995. Regulated streamﬂow, ﬁne-grained deposits and effective discharge in canyons with abundant debris fans. In: Costa, J.E., Miller, A.J., Potter, K.W., and Wilcock, P.R. (Eds), Natural and Anthropogenic Inﬂuences in Fluvial Geomorphology. American Geophysical Union Press, Washington, DC, pp. 177–195, Geophysical Monograph 89. Scott, K.M., Gravlee, G.C., 1968. Flood surge on the Rubicon River, California – hydrology, hydraulics, and boulder transport. US Geol. Surv. Prof. Pap. 422-M, 38. Scott, M.L., Friedman, J.M., Auble, G.T., 1996. Fluvial process and the establishment of bottomland trees. Geomorphology 14, 327–339. Sear, D.A., 1994. River restoration and geomorphology. Aquat. Conserv. Mar. Freshwat. Ecosyst. 4, 169–177. Seidl, M.A., Dietrich, W.E., 1992. The problem of channel erosion into bedrock. Catena Suppl. 23, 101–124. Seidl, M.A., Dietrich, W.E., Kirchner, J.W., 1994. Longitudinal proﬁle development into bedrock: An analysis of Hawaiian channels. J. Geol. 102, 457–474. Shroba, R.R., Schmidt, P.W., Crosby, E.J., Hansen, W.R., Soule, J.M., 1979. Storm and ﬂood of July 31–August 1, 1976, in the Big Thompson River and Cache la Poudre River basins, Larimer and Weld Counties, Colorado. US Geol. Surv. Prof. Pap. 1115, 152. Shroder, J.F., Bishop, M.P., Scheppy, R., 1998. Catastrophic ﬂood ﬂushing of sediment, western Himalaya, Pakistan. In: Kalvoda, J. and Rosenfeld, C.L. (Eds), Geomorphological Hazards in High Mountain Areas. Kluwer Academic Publishers, Dordrecht, pp. 27–48. Shvidchenko, A.B., Pender, G., Hoey, T.B., 2001. Critical shear stress for incipient motion of sand/gravel streambeds. Water Resour. Res. 37, 2273–2283. Simons, D.B., Richardson, E.V., 1966. Resistance to ﬂow in alluvial channels. US Geol. Surv. Prof. Pap. 422J, 61. Sklar, L., Dietrich, W.E., 1998. River longitudinal proﬁles and bedrock incision models: Stream power and the inﬂuence of sediment supply. In: Tinkler, K.J. and Wohl, E.E. (Eds), Rivers Over Rock: Fluvial Processes in Bedrock Channels. American Geophysical Union Press, Washington, DC, pp. 237–260, Geophysical Monograph 107. Sklar, L.S., Dietrich, W.E., 2001. Sediment and rock strength controls on river incision into bedrock. Geology 29, 1087–1090. Sklar, L.S., Dietrich, W.E., 2004. A mechanistic model for river incision into bedrock by saltating bed load. Water Resour. Res. 40, W06301. Sloan, J., Miller, J.R., Lancaster, N., 2001. Response and recovery of the Eel River, California, and its tributaries to ﬂoods in 1955, 1964, and 1997. Geomorphology 36, 129–154. Smith, J.D., 2004. The role of riparian shrubs in preventing ﬂoodplain unraveling along the Clark Fork of the Columbia River in the Deer Lodge Valley, Montana. In: Bennett, S.J. and Simon, A. (Eds), Riparian Vegetation and Fluvial Geomorphology. American Geophysical Union Press, Washington, DC, pp. 71–85. Smith, R.D., Sidle, R.C., Porter, P.E., 1993. Effects on bedload transport of experimental removal of woody debris from a forest gravel-bed stream. Earth Surf. Process. Landf. 18, 455–468. Sparks, R.E., Bayley, P.B., Kohler, S.L., Osborne, L.L., 1990. Disturbance and recovery of large ﬂoodplain rivers. Environ. Manage. 14, 699–709. Springer, G.S., Wohl, E.E., 2002. Empirical and theoretical investigations of sculpted forms in Buckeye Creek Cave, West Virginia. J. Geol. 110, 469–481. Stanford, J.A., Ward, J.V., Liss, W.J., et al., 1996. A general protocol for restoration of regulated rivers. Regul. Rivers Res. Manage. 12, 391–413. Starkel, L., 1988. Tectonic, anthropogenic and climatic factors in the history of the Vistula River valley downstream of Cracow. In: Lang, G. and Schlu¨chter, C. (Eds), Lake, Mire and River Environments. Balkema, Rotterdam, pp. 161–170.

210

E. Wohl

Stewart, J.H., LaMarche, V.C., 1967. Erosion and deposition produced by the ﬂood of December 1964 on Coffee Creek, Trinity County, California. US Geol. Surv. Prof. Pap. 422-K, 22. Stock, J., Dietrich, W.E., 2003. Valley incision by debris ﬂows: Evidence of a topographic signature. Water Resour. Res. 39, ESG 1-1 to 1-25. Stover, S.C., Montgomery, D.R., 2001. Channel change and ﬂooding, Skokomish River, Washington. J. Hydrol. 243, 272–286. Sturdevant-Rees, P., Smith, J.A., Morrison, J., Baeck, M.L., 2001. Tropical storms and ﬂood hydrology of the central Appalachians. Water Resour. Res. 37, 2143–2168. Surian, N., Andrews, E.D., 1999. Estimation of geomorphically signiﬁcant ﬂows in alpine streams of the Rocky Mountains, Colorado (USA). Regul. Rivers Res. Manage. 15, 273–288. Sutherland, D.G., Ball, M.H., Hilton, S.J., Lisle, T.E., 2002. Evolution of a landslide-induced sediment wave in the Navarro River, California. Geol. Soc. Am. Bull. 114, 1036–1048. Swanson, F.J., Johnson, S.L., Gregory, S.V., Acker, S.A., 1998. Flood disturbance in a forested mountain landscape. BioScience 48, 681–689. Tal, M., Gran, K., Murray, A.B., et al., 2004. Riparian vegetation as a primary control on channel characteristics in multi-thread rivers. In: Bennett, S.J. and Simon, A. (Eds), Riparian Vegetation and Fluvial Geomorphology. American Geophysical Union Press, Washington, DC, pp. 43–58. Thompson, D.M., 1995. The effects of large organic debris on sediment processes and stream morphology in Vermont. Geomorphology 11, 235–244. Tianche, L., Schuster, R.L., Wu, J., 1986. Landslide dams in south-central China. In: Schuster, R.L. (Ed.), Landslide Dams: Processes, Risk, and Mitigation. American Society of Civil Engineers, New York, pp. 146–162. Tinkler, K.J., 1971. Active valley meanders in south-central Texas and their wider implications. Geol. Soc. Am. Bull. 82, 1783–1800. Tinkler, K.J., 1993. Fluvially sculpted rock bedforms in Twenty Mile Creek, Niagara Peninsula, Ontario. Can. J. Earth Sci. 30, 945–953. Tinkler, K.J., Wohl, E.E. (Eds), 1998. Rivers Over Rock: Fluvial Processes in Bedrock Channels. American Geophysical Union Press, Washington, DC, Geophysical Monograph 107. Tooth, S., 2000. Process, form and change in dryland rivers: A review of recent research. Earth Sci. Rev. 51, 67–107. Van Rijn, L.C., 1984. Sediment transport, Part III: Bed forms and alluvial roughness. J. Hydraul. Eng. 110, 1733–1754. Warburton, J., 1994. Channel change in relation to meltwater ﬂooding, Bas Glacier d’Arolla, Switzerland. Geomorphology 11, 141–149. Ward, J.V., Tockner, K., Uehlinger, U., Malard, F., 2001. Understanding natural patterns and processes in river corridors as the basis for effective river restoration. Regul. Rivers Res. Manage. 17, 311–323. Webb, R.H. and Betancourt, J.L., 1990. Climatic variability and ﬂood frequency of the Santa Cruz River, Pima County, Arizona. US Geol. Surv. Open-File Report 90-553, 69pp. Webb, R.H., Pringle, P.T., Rink, G.R., 1989. Debris ﬂows from tributaries of the Colorado River, Grand Canyon National Park, Arizona. US Geol. Surv. Prof. Pap. 1492, 39. Wende, R., Nanson, G.C., 1998. Anabranching rivers: Ridge-form alluvial channels in tropical northern Australia. Geomorphology 22, 205–224. Wiele, S.M., Graf, J.B., Smith, J.D., 1996. Sand deposition in the Colorado River in the Grand Canyon from ﬂooding of the Little Colorado River. Water Resour. Res. 32, 3579–3596. Williams, G.P., Guy, H.P., 1973. Erosional and depositional aspects of Hurricane Camille in Virginia, 1969. US Geol. Surv. Prof. Pap. 804, 80. Williams, G.P. and Wolman, M.G., 1985. Effects of dams and reservoirs on surface-water hydrology – changes in rivers downstream from dams. U.S. Geological Survey National Water Summary 1985, pp. 83–88. Wohl, E.E., 1992a. Bedrock benches and boulder bars: Floods in the Burdekin Gorge of Australia. Geol. Soc. Am. Bull. 104, 770–778. Wohl, E.E., 1992b. Gradient irregularity in the Herbert Gorge of northeastern Australia. Earth Surf. Process. Landf. 17, 69–84. Wohl, E.E., 1993. Bedrock channel incision along Piccaninny Creek, Australia. J. Geol. 101, 749–761.

Review of effects of large floods in resistant-boundary channels

211

Wohl, E., 1998. Bedrock channel morphology in relation to erosional processes. In: Tinkler, K.J. and Wohl, E.E. (Eds), Rivers Over Rock: Fluvial Processes in Bedrock Channels. American Geophysical Union Press, Washington, DC, pp. 133–151, Geophysical Monograph 107. Wohl, E.E., 2000a. Anthropogenic impacts on ﬂood hazards. In: Wohl, E.E. (Ed.), Inland Flood Hazards: Human, Riparian, and Aquatic Communities. Cambridge University Press, Cambridge, UK. Wohl, E.E., 2000b. Mountain rivers. American Geophysical Union Press, Washington, DC, 320pp. Wohl, E.E., Angermeier, P.L., Bledsoe, B., Kondolf, G.M., MacDonnell, L., Merritt, D.M., Palmer, M.A., Poff, N.L. and Tarboton, D., 2005. River restoration. Water Resour. Res. 41, W10301. Wohl, E.E., Cenderelli, D.A., 2000. Sediment deposition and transport patterns following a reservoir sediment release. Water Resour. Res. 36, 319–333. Wohl, E., Cenderelli, D., Mejia-Navarro, M., 2001. Channel change from extreme ﬂoods in canyon rivers. In: Anthony, D.J., Harvey, M.D., Laronne, J.B., and Mosley, M.P. (Eds), Applying Geomorphology to Environmental Management. Water Resources Publications, Highlands Ranch, CO, pp. 149–174. Wohl, E.E., Pearthree, P.A., 1991. Debris ﬂows as geomorphic agents in the Huachuca Mountains of southeastern Arizona. Geomorphology 4, 273–292. Wohl, E., Springer, G., 2005. Bedrock channel incision along the upper Rio Chagres basin, Panama. In: Harmon, R.S. (Ed.), The Rio Chagres: A multidisciplinary proﬁle of a tropical watershed. Springer, Dodrecht, The Netherlands, pp. 189–209. Wohl, E.E., Thompson, D.M., Miller, A.J., 1999. Canyons with undulating walls. Geol. Soc. Am. Bull. 111, 949–959. Wolman, M.G., 1967. A cycle of sedimentation and erosion in urban river channels. Geogr. Ann. 49A, 385–395. Wolman, M.G., Miller, J.P., 1960. Magnitude and frequency of forces in geomorphic processes. J. Geol. 68, 54–74. Wondzell, S.M., Swanson, F.J., 1999. Floods, channel change, and the hyporheic zone. Water Resour. Res. 35, 555–567. Wroblicky, G.J., Campana, M.E., Valett, H.M., Dahm, C.N., 1998. Seasonal variation in surfacesubsurface water exchange and lateral hyporheic area of two stream-aquifer systems. Water Resour. Res. 34, 317–328. Wyzga, B., 1996. Changes in the magnitude and transformation of ﬂood waves subsequent to the channelization of the Raba River, Polish Carpathians. Earth Surf. Process. Landf. 21, 749–763. Young, W.J., Olley, J.M., Prosser, I.P., Warner, R.F., 2001. Relative changes in sediment supply and sediment transport capacity in a bedrock-controlled river. Water Resour. Res. 37, 3307–3320. Zen, E., Prestegaard, K.L., 1994. Possible hydraulic signiﬁcance of two kinds of potholes: Examples from the paleo-Potomac River. Geology 22, 47–50. Zielinski, T., 2003. Catastrophic ﬂood effects in alpine/foothill ﬂuvial system (a case study from the Sudetes Mts, SW Poland). Geomorphology 54, 293–306. Zimmermann, A., Church, M., 2001. Channel morphology, gradient proﬁles, and bed stresses during ﬂood in a step-pool channel. Geomorphology 40, 311–327.

Discussion by G. Heritage and D. Milan E. Wohl notes the importance of large ﬂoods on inﬂuencing the geomorphology of resistant-boundary channels. The authors have also found this to be true for the bedrock inﬂuenced semiarid Sabie River, South Africa. Analysis of 1:10,000 scale aerial photographs of the morphologic response of the river to low, moderate, and extreme ﬂows has revealed a complex response across all ﬂows that may be rationalized when viewed at the reach and morphologic unit scale. Bedrock dominated channel types display little change in response to low to moderate ﬂows whereas alluviated channel reaches change was variable in direction and degree. At the level

Review of effects of large floods in resistant-boundary channels

211

Wohl, E., 1998. Bedrock channel morphology in relation to erosional processes. In: Tinkler, K.J. and Wohl, E.E. (Eds), Rivers Over Rock: Fluvial Processes in Bedrock Channels. American Geophysical Union Press, Washington, DC, pp. 133–151, Geophysical Monograph 107. Wohl, E.E., 2000a. Anthropogenic impacts on ﬂood hazards. In: Wohl, E.E. (Ed.), Inland Flood Hazards: Human, Riparian, and Aquatic Communities. Cambridge University Press, Cambridge, UK. Wohl, E.E., 2000b. Mountain rivers. American Geophysical Union Press, Washington, DC, 320pp. Wohl, E.E., Angermeier, P.L., Bledsoe, B., Kondolf, G.M., MacDonnell, L., Merritt, D.M., Palmer, M.A., Poff, N.L. and Tarboton, D., 2005. River restoration. Water Resour. Res. 41, W10301. Wohl, E.E., Cenderelli, D.A., 2000. Sediment deposition and transport patterns following a reservoir sediment release. Water Resour. Res. 36, 319–333. Wohl, E., Cenderelli, D., Mejia-Navarro, M., 2001. Channel change from extreme ﬂoods in canyon rivers. In: Anthony, D.J., Harvey, M.D., Laronne, J.B., and Mosley, M.P. (Eds), Applying Geomorphology to Environmental Management. Water Resources Publications, Highlands Ranch, CO, pp. 149–174. Wohl, E.E., Pearthree, P.A., 1991. Debris ﬂows as geomorphic agents in the Huachuca Mountains of southeastern Arizona. Geomorphology 4, 273–292. Wohl, E., Springer, G., 2005. Bedrock channel incision along the upper Rio Chagres basin, Panama. In: Harmon, R.S. (Ed.), The Rio Chagres: A multidisciplinary proﬁle of a tropical watershed. Springer, Dodrecht, The Netherlands, pp. 189–209. Wohl, E.E., Thompson, D.M., Miller, A.J., 1999. Canyons with undulating walls. Geol. Soc. Am. Bull. 111, 949–959. Wolman, M.G., 1967. A cycle of sedimentation and erosion in urban river channels. Geogr. Ann. 49A, 385–395. Wolman, M.G., Miller, J.P., 1960. Magnitude and frequency of forces in geomorphic processes. J. Geol. 68, 54–74. Wondzell, S.M., Swanson, F.J., 1999. Floods, channel change, and the hyporheic zone. Water Resour. Res. 35, 555–567. Wroblicky, G.J., Campana, M.E., Valett, H.M., Dahm, C.N., 1998. Seasonal variation in surfacesubsurface water exchange and lateral hyporheic area of two stream-aquifer systems. Water Resour. Res. 34, 317–328. Wyzga, B., 1996. Changes in the magnitude and transformation of ﬂood waves subsequent to the channelization of the Raba River, Polish Carpathians. Earth Surf. Process. Landf. 21, 749–763. Young, W.J., Olley, J.M., Prosser, I.P., Warner, R.F., 2001. Relative changes in sediment supply and sediment transport capacity in a bedrock-controlled river. Water Resour. Res. 37, 3307–3320. Zen, E., Prestegaard, K.L., 1994. Possible hydraulic signiﬁcance of two kinds of potholes: Examples from the paleo-Potomac River. Geology 22, 47–50. Zielinski, T., 2003. Catastrophic ﬂood effects in alpine/foothill ﬂuvial system (a case study from the Sudetes Mts, SW Poland). Geomorphology 54, 293–306. Zimmermann, A., Church, M., 2001. Channel morphology, gradient proﬁles, and bed stresses during ﬂood in a step-pool channel. Geomorphology 40, 311–327.

Discussion by G. Heritage and D. Milan E. Wohl notes the importance of large ﬂoods on inﬂuencing the geomorphology of resistant-boundary channels. The authors have also found this to be true for the bedrock inﬂuenced semiarid Sabie River, South Africa. Analysis of 1:10,000 scale aerial photographs of the morphologic response of the river to low, moderate, and extreme ﬂows has revealed a complex response across all ﬂows that may be rationalized when viewed at the reach and morphologic unit scale. Bedrock dominated channel types display little change in response to low to moderate ﬂows whereas alluviated channel reaches change was variable in direction and degree. At the level

212

E. Wohl

of morphologic unit changes were most dramatic for unconsolidated bar deposits, which were eroded and re-deposited frequently. Larger macrochannel features remained static and cohesive bars displayed sporadic vertical accretion in response to moderate ﬂoods. Morphologic change is thus generally restricted to features associated with the base of the incised river, particularly those composed of unconsolidated sediment or those located close to high energy areas of the channel. The response to extreme ﬂows was more dramatic with signiﬁcant losses of cohesive deposits occurring in the base of the incised channel. Bedrock dominated channel reaches remained largely unchanged, whereas alluviated channel reaches experienced both erosion and deposition. Morphologic unit change was widespread with unconsolidated active channel bar deposits eroded. Macrochannel deposits appear largely unaffected by the extreme ﬂood magnitudes associated with the contemporary ﬂow regime. Such changes illustrate the variable response to ﬂoods of varying magnitude exhibited by the different morphologic units found in the Sabie River. More importantly, it also demonstrates the variable recovery rates seen between unit types. These response–recovery patterns have been summarized as a series of morphologic response models (Fig. 8.4). It is suggested that these would be useful to river managers as well as scientists who often need to place into context the apparently dramatic changes that occur to semiarid river systems subject to extreme ﬂow regimes.

Figure 8.4. Morphological response models for the Sabie River, South Africa.

Gravel-Bed Rivers VI: From Process Understanding to River Restoration H. Habersack, H. Pie´gay, M. Rinaldi, Editors r 2008 Elsevier B.V. All rights reserved.

213

9 Modelling river-bank-erosion processes and mass failure mechanisms: progress towards fully coupled simulations Massimo Rinaldi and Stephen E. Darby

Abstract This paper reviews recent developments in modelling the two main sets of bankerosion processes and mechanisms, namely ﬂuvial erosion and mass failure, before suggesting an avenue for research to make further progress in the future. Our review of mass failure mechanisms reveals that the traditional use of limit equilibrium methods to analyse bank stability has in recent years been supplemented by research that has made progress in understanding and modelling the role of positive and negative pore water pressures, conﬁning river pressures, and hydrograph characteristics. While understanding of both ﬂuvial erosion and mass failure processes has improved in recent years, we identify a key limitation in that few studies have examined the nature of the interaction between these processes. We argue that such interactions are likely to be important in gravel-bed rivers and present new simulations in which ﬂuvial erosion, pore water pressure, and limit equilibrium stability models are combined into a fully coupled analysis. The results suggest that existing conceptual models, which describe how bank materials are delivered to the ﬂuvial sediment transfer system, may need to be revised to account for the unforeseen effects introduced by feedback between the interacting processes. 1.

Introduction

Bank erosion is a key process in ﬂuvial dynamics, affecting a wide range of physical, ecological, and socio-economic issues in the ﬂuvial environment. These include the establishment and evolution of river and ﬂoodplain morphology and their associated habitats (e.g., Hooke, 1980; Millar and Quick, 1993; Darby and Thorne, 1996a; Barker et al., 1997; Millar, 2000; Goodson et al., 2002), turbidity problems (e.g., Bull, 1997; Eaton et al., 2004), sediment, nutrient, and contaminant dynamics (e.g., Reneau et al., 2004), loss of riparian lands (e.g., Amiri-Tokaldany et al., 2003), and associated threats to ﬂood defence and transportation infrastructure (e.g., Simon, E-mail address: [email protected]ﬁ.it (M. Rinaldi) ISSN: 0928-2025

DOI: 10.1016/S0928-2025(07)11126-3

M. Rinaldi, S.E. Darby

214

1995). Moreover, recent studies have shown that the contribution of bank-derived sediments to catchment sediment budgets may be higher than previously thought, although the precise fraction varies depending on the time-scale of measurement (Bull, 1997). For example, considering annual sediment yields, Walling et al. (1999) showed that bank sediments contribute up to 37% of the total (10,816 t/yr) suspended sediment yield, even in the relatively low-energy catchments of the UK, with the contribution rising to values as high as 80% of the total (75,000 t/yr) suspended sediment yield in some highly unstable, incised, channel systems (e.g., Simon and Darby, 2002). With such a signiﬁcant fraction of material within the alluvial sediment system derived from river banks, it is evident that knowledge of the rates, patterns, and controls on bank-erosion events that release sediment to river systems is a pre-requisite for a complete understanding of the ﬂuvial sediment transport regime. Naturally, much research has already been devoted to these issues. These contributions include a number of excellent reviews (Lawler, 1993; Lawler et al., 1997b; Couper, 2004), including those published by Grissinger (1982) and Thorne (1982) in the original Gravel-Bed Rivers volume (Hey et al., 1982). So what might ‘yet’ another review of bank-erosion processes actually achieve? As Fig. 9.1 shows, there is a growing number of bank-erosion investigations (38% of the publications appear after 1997) and a shift in the pattern of ‘hot’ topics in the discipline. In particular, new research has elucidated the role of riparian vegetation (e.g., Abernethy and 18 16

interaction

Number of papers

14

others

12

vegetation

10

stability

8

erosion

GBR1

6 4 2 2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

1977

1975

1973

1971

1969

1967

1965

1963

1961

1959

0

Figure 9.1. Summary of the bibliographic review on river-bank-erosion processes conducted for this chapter (total of 194 papers considered). Erosion: papers focused on ﬂuvial entrainment; stability: papers on mass failures and bank stability; vegetation: papers focusing on the role of vegetation; others: papers on other issues related to bank erosion (e.g., measurement of bank retreat, variables controlling rates of retreat, sediment delivery from bank processes, inﬂuence of bank processes on channel geometry, etc.); interaction: papers on modelling width adjustments and channel migration, and including to some extent the interaction between ﬂuvial erosion and mass failures. Dates of major reviews of bank erosion in the ﬁrst Gravel-Bed Rivers I proceedings volume in 1982 (GBR I) and the most recent major review in 1997 (dashed line) are also highlighted.

Modelling river-bank-erosion processes and mass failure mechanisms

215

Rutherfurd, 1998, 2000; Simon and Collison, 2002) and bank hydrology (e.g., Rinaldi and Casagli, 1999; Casagli et al., 1999; Rinaldi et al., 2004) as key controlling inﬂuences on bank stability. In contrast, although improvements in the modelling of near-bank ﬂows are starting to be made (e.g., Kean and Smith, 2006a,b), there are still relatively few studies that have been concerned with the process of ﬂuvial erosion (i.e., the removal of bank sediments by the direct action of the ﬂow). Accordingly, little progress has been made in understanding ﬂuvial bank erosion of cohesive sediments since the contributions of Arulanandan et al. (1980) and Grissinger (1982). Notable exceptions to this trend include some work that has sought to quantify entrainment thresholds and process rates (e.g., Lawler et al., 1997a; Simon et al., 2000; Dapporto, 2001). The role of weathering as a signiﬁcant agent of erosion has also started to be recognised (e.g., Lawler, 1993; Prosser et al., 2000; Couper and Maddock, 2001), both in headwater reaches (where weathering may be the dominant mechanism by which sediment is removed from the bank face) and, elsewhere, as a mechanism for enhancing bank erodibility and promoting ﬂuvial erosion. Fig. 9.1 also highlights another gap in the literature. While most studies recognise that bank retreat is the integrated product of three interacting processes (namely, weathering and weakening, ﬂuvial erosion, and mass failure), they tend to adopt reductionist approaches that focus on a single set of processes, so interactions between different groups of processes are not usually considered. This is important because dynamic interactions and feedbacks between processes may lead to outcomes that are not predictable a priori. In short, viewing bank processes in isolation is unrealistic and introduces the possibility that conclusions derived from such studies are biased. Recognition of this problem is not new. Lawler (1992) introduced a conceptual model of changing bank process dominance in a hypothetical drainage basin (Fig. 9.2), emphasising that processes act not in isolation, but are always present to a varying degree. While Fig. 9.2 represents a conceptualisation of an idealised basin and the length scales therein are, therefore, deliberately omitted, it is instructive to attempt to contextualise the drainage basin locations within which recent bankerosion research has been conducted. Bearing in mind that these studies have typically sought to isolate the effects of individual process groups, it is noteworthy that they cluster in the mid- to downstream reaches, where process interactions are strongest. Interactions between mass failure and ﬂuvial-erosion processes (as opposed to the role of individual processes acting in isolation) therefore have particular relevance in the context of gravel-bed rivers, as the zone of interaction coincides at least in part with the middle reaches of basins where gravel-beds are typical, and also because the dominance of subaerial processes is generally limited geographically to the headwaters of typical ﬂuvial systems (Couper and Maddock, 2001). This paper therefore seeks to address two objectives. First, we review recent developments regarding the two main bank-erosion phenomena (ﬂuvial erosion and mass failure) responsible for bank retreat in gravel-bed rivers. Second, we focus on studies which have sought to address the interactions between these two processes and mechanisms. Included in this synthesis are new ﬁndings from our own research which show that adopting a fully coupled modelling approach that views bank processes as interacting, rather than individual, entities leads to a distinctive vision of

M. Rinaldi, S.E. Darby

216

Effectiveness of process grousp

G C

S

A

Fluid entrainment

Preparation processes F re

eze

/th a

Mass failure

w

dessication

Distance downstream Figure 9.2. Conceptual model of process dominance in the ﬂuvial system. (After Lawler (1992)). Also shown are the approximate locations within their respective basins (mapped as proportion of stream length) of the sites used in some recent bank-erosion studies: A ¼ River Asker, Dorset, UK; C ¼ Cecina River, Tuscany, Italy; G ¼ Goodwin Creek, Mississippi, USA; S ¼ Sieve River, Tuscany, Italy. (Reproduced with permission from Wiley and Sons, 1992.)

the ways in which bank-derived materials are delivered to the alluvial sediment transfer system.

2.

Modelling ﬂuvial erosion

Fluvial erosion is deﬁned as the removal of bank material by the action of hydraulic forces, although it generally occurs in combination with weathering processes that prepare bank sediments for erosion by enhancing their erodibility (Hooke, 1980; Thorne, 1982; Lawler, 1993; ASCE Task Committee on Hydraulics, Bank Mechanics, and Modeling of River Width Adjustment, 1998; Prosser et al., 2000; Couper and Maddock, 2001). Relative to mass failure, ﬂuvial erosion is, at the scale of the ﬂow event and once the critical entrainment threshold has been exceeded, a quasi-continuous process, with the volume of sediment delivered by ﬂuvial erosion dependent on the duration of the competent ﬂow. In general, ﬂuvial-erosion rates depend on the near-bank ﬂow intensity and physical characteristics (i.e., the erodibility) of the bank material. However, this simple conceptualisation masks enormous complexity that results from the inherent variability of the relevant controlling parameters. Thus, observed rates of ﬂuvial-erosion range over several orders of magnitude (Hooke, 1980) and ﬂuvial-erosion rates are predictable only to the extent that the controlling parameter values, and their inherent variability, can be estimated accurately.

Modelling river-bank-erosion processes and mass failure mechanisms

217

It is widely accepted that the rate of ﬂuvial bank erosion can be quantiﬁed using an excess shear stress formula such as (Partheniades, 1965; Arulanandan et al., 1980): ¼ kd ðt tc Þa

(9.1)

where e (m/s) is the ﬂuvial bank-erosion rate per unit time and unit bank area, t (Pa) is the boundary shear stress applied by the ﬂow, kd (m2s/kg) and tc (Pa) are erodibility parameters (erodibility coefﬁcient, kd, and critical shear stress, tc), and a (dimensionless) is an empirically derived exponent. It is important to note that although excess shear stress models of this type are widely accepted and used in a range of geomorphological applications (e.g., Arulanandan et al., 1980; Govers, 1991; Howard, 1994), no formal validation of this model has yet been undertaken. Thus some uncertainty remains over the value of the exponent a (which is commonly assumed to take a value close to 1 for most studies involving cohesive sediments, e.g., Partheniades, 1965). Perhaps more signiﬁcantly, the physical basis of the excess shear stress model for bank erosion can be questioned. One problem is its reliance on a threshold value, which is difﬁcult to incorporate into numerical models due to the sharp threshold between stability and failure, which in turn results in instabilities near the threshold value. Nevertheless, such threshold behaviour is appropriate, particularly on cohesive river banks. Any propagation of numerical error may, therefore, de facto require the erodibility coefﬁcient (kd) to be treated as a calibration parameter, a problem highlighted recently by Crosato (2007). For the purposes of this review we assume that the basic form of equation (9.1) is robust and that predictive ability is limited by the need to estimate the necessary parameter values accurately. In subsequent sub-sections we therefore focus on recent developments concerned with improving estimates of the erosion rate, erodibility, and shear stress parameters.

2.1.

Erosion rate

A comprehensive review of the methods used to observe bank erosion was provided by Lawler (1993). Recently, techniques such as digital photogrammetry and laser scanning (e.g., Lane et al., 1994; Barker et al., 1997; Nagihara et al., 2004) can provide the opportunity to deﬁne river bank topography at unprecedented spatial resolution (surveys with point densities of ca. 107 points across a bank face are readily obtainable using terrestrial laser scanning) and accuracy (72 mm). Bank erosion can then be quantiﬁed using the survey data to construct Digital Terrain Models (DTMs) for time intervals and differencing to establish net change. However, logistical and safety concerns usually limit the frequency of monitoring to relatively coarse timescales, at best perhaps resolving individual ﬂow events. This is problematic because the pre- versus post-flow event ‘window’ is not the same thing as the bank erosion event window, such that it is not usually possible to resolve process thresholds, timing, and rates (Lawler, 2005). To address this limitation, new quasi-continuous bank-erosion sensors based on the use of photoelectronic cells (PEEPs; e.g., Lawler, 1993; Lawler et al., 1997a) and thermal consonance timing (TCT; e.g., Lawler, 2005) have been

M. Rinaldi, S.E. Darby

218

developed, though they have not yet been widely deployed. While these approaches are promising, the use of sensors can disturb the bank face, while excellent temporal resolution is inevitably obtained at relatively low spatial resolution. While accurate and representative discrimination of bank-erosion rates therefore remains elusive, studies that combine high spatial/low temporal (e.g., photogrammetry) and high temporal/low spatial (e.g., PEEPs, TCT) resolution approaches may deliver exciting new results in the near future.

2.2.

Erodibility of bank sediment

For granular (non-cohesive) sediments, bank erodibility parameters are modelled based on the same methods that are used to predict the entrainment of bed sediments, albeit with modiﬁcations to take into account the effect of the bank angle on the downslope component of the particle weight (Lane, 1955) and the case of partly packed and cemented sediments (e.g., Millar and Quick, 1993; Millar, 2000). Determination of critical shear stress for cohesive materials is more complex, given that it is widely recognised that ﬂuvial entrainment for cohesive sediments depends on several factors, including (amongst others) clay and organic content, and the composition of interstitial ﬂuids (Arulanandan et al., 1980; Grissinger, 1982; Knapen et al., 2007). Consequently methods for predicting the erodibility of cohesive banks remain poor. To address this issue, recent studies have deployed in situ jet-testing devices (e.g., Hanson, 1990; Hanson and Simon, 2001) to obtain direct measurements of bank erodibility (e.g., Dapporto, 2001). This is achieved by directing a jet of water with known hydraulic properties at the bank material. The resulting deformation is measured periodically with a mechanical point gauge, until an equilibrium scour depth is attained. The measured deformation rate, scour depth, and known hydraulic properties are used to determine the erodibility parameters. While jet-testers offer in situ sampling, our experience is that their design (especially their large weight) makes their deployment to inaccessible sites difﬁcult, and it is also hard to emplace them without disturbing the bank surface. Moreover, individual tests are time consuming (ca. 0.5 h), making it difﬁcult to obtain the numbers of samples needed to adequately characterise the spatial and temporal variability of the bank materials. On resistant surfaces, errors involved in mechanically inserting the point gauge into the base of the scour hole can be similar in magnitude to the scour depth itself, while erodible materials generate scour depths that can exceed the extent of the gauge. Instruments such as the Cohesive Strength Meter (Tolhurst et al., 1999) appear to offer advantages over conventional jet-testing devices. The CSM is similar to these in that water jets of increasing strength are directed at the target surface. However, instead of measuring the resulting scour depth, the CSM detects erosion by monitoring optical transmission in an enclosed sampling head chamber. Thus, the moment of erosion corresponds to sudden reductions in optical transmission induced by the suspension of eroded sediment within the test chamber, with the jet properties at that threshold deﬁning the critical stress. Tests are both automated and rapid (o3 min) so the device can easily be used to obtain large numbers of samples. So far

Modelling river-bank-erosion processes and mass failure mechanisms

219

it has only been deployed in estuarine environments (Tolhurst et al., 1999), but the CSM appears to offer a potentially fruitful avenue of bank-erosion research.

2.3.

Near-bank shear stresses

With the aforementioned recent developments in bank-erosion monitoring technology and in the quantiﬁcation of bank erodibility, the ‘missing link’ in equation (9.1) remains the difﬁculty of characterising the ﬂuid stresses that are exerted on river banks during the large ﬂows that typically drive erosion. Bank boundary shear stress is highly variable both in space and time, dependent as it is on such factors as the bank geometry (which is itself highly variable), cross-section size and shape, channel curvature, and ﬂow stage. This variability presents a challenge for anyone seeking to characterise the shear stress distribution via direct measurement. Sampling strategies would need to capture this natural variability, suggesting that the necessary ﬂow velocimetry equipment would need to be deployed at high spatial and temporal resolution, during the (hazardous) high ﬂow conditions associated with bank erosion. It is, therefore, unsurprising that such investigations are lacking, apart from some ﬂume studies (e.g., Blanckaert and Graf, 2001) which are able to achieve the necessary sampling resolution, albeit under rather idealised conditions. If ﬁeld data collection is impractical, the only viable alternative is to predict the shear stress values using hydraulic models. Some models have been developed using empirical data sets obtained from laboratory channels (Leutheusser, 1963; Kartha and Leutheusser, 1972; Simons and Sentu¨rk, 1977; Knight et al., 1984), but these can only be applied with caution to natural rivers, as the bank and channel forms present in ﬂumes with regular geometry represent the problem rather poorly. Recently, progress has been made in using analytical models to quantify form roughness induced by the irregular bank morphology and partition the shear stress acting on the banks (e.g., Kean and Smith, 2004, 2006a,b; Grifﬁn et al., 2005). Although these approaches are promising, it is not yet clear whether such approaches are entirely appropriate. Speciﬁcally, a lack of ﬁeld data sets means that we simply do not yet know whether near-bank ﬂows are dominated by the form drag induced by the topographic irregularities (e.g., embayments, slump deposits, etc.) associated with natural, eroding, banks (e.g., Thorne and Furbish, 1995), or by the effects of turbulence induced by strong lateral shear and the occurrence of wakes. If the latter is the case, then modelling near-bank ﬂows would require the application of 3-dimensional Computational Fluid Dynamics (3D-CFD) modelling techniques. The practice of using 3D-CFD modelling techniques as a substitute for ﬁeld data in river ﬂows that are difﬁcult or impossible to measure has now become established for a range of open-channel ﬂow contexts (e.g., Nicholas and Walling, 1997, 1998; Hodskinson and Ferguson, 1998; Nicholas and Sambrook-Smith, 1999; Bradbrook et al., 2000; Lane et al., 2000; Darby et al., 2004). However, the application of 3DCFD to near-bank ﬂows remains novel and replication of near-bank ﬂows would depend on: (i) ensuring the discretised computational scheme accurately solves the underlying conservation equations; (ii) selecting an appropriate turbulence-closure model (TCM), and (iii) accurately deﬁning the initial and boundary conditions

M. Rinaldi, S.E. Darby

220

(Darby et al., 2004). Accepted standards of computational mesh design (e.g., American Society of Mechanical Engineers (ASME), 1993; American Institute of Aeronautics and Astronautics (AIAA), 1998) are already available, with adherence to those standards ensuring that discretisation and numerical solution errors are minimised (Hardy et al., 2003). Consequently, we focus attention on the latter two issues, though we note that very high-resolution grids are likely to be needed to represent the complex ﬂow structures in near-bank environments. This, in turn, may create additional problems, not only of large computational requirements, but in terms of deﬁning the boundary conditions correctly. Regarding the parameterisation of turbulence, some studies have begun to investigate the potential for approaches such as Direct Numerical Simulation (DNS) and Large Eddy Simulation (LES) to deliver accurate hydraulic data (e.g., Rodi et al., 1997). However, even with the capabilities of modern computing, these approaches have only been applied to ﬂows with ﬁxed (non-deformable) channel boundaries. For morphological modelling, Reynolds averaging appears set to continue as the only feasible approach, at least for the foreseeable future. However, all such TCMs contain empirical elements, so model selection must be matched to the anticipated physical conditions, namely: (i) strong lateral shear; (ii) occurrence of separated ﬂow around topographic irregularities (e.g., embayments, slump deposits, etc.) associated with eroding banks. These requirements suggests that an anisotropic TCM is required (e.g., So et al., 1993; Sotiropoulos, 2001; Blanckaert and Graf, 2001; Gerolymos et al., 2002), and a Reynolds Stress Model (RSM) appears most appropriate for the speciﬁc context of modelling near-bank ﬂows.

3.

Modelling river bank failures

Mass failure is the collapse and movement of bank material under gravity. Relative to ﬂuvial erosion, mass failure is discontinuous and large-scale and occurs by any of a number of mechanisms (Thorne, 1982), with a speciﬁc model required for each. The methods developed and used in the literature have concentrated only on a relatively few of these, in particular slides (planar or rotational) and cantilever failures. Referring to the classical mechanism of a planar slide (Lohnes and Handy, 1968), bank failure occurs when the destabilising forces, due to gravity, exceed the resisting forces, which are related to the shear strength of the bank materials expressed by the failure criterion of Fredlund et al. (1978) as: t ¼ c0 þ ðs ua Þ tan f0 þ ðua uw Þ tan fb

(9.2)

where t is the shear strength (kPa), c0 the effective cohesion (kPa), s the normal stress (kPa), ua the pore air pressure (kPa), f0 the friction angle in terms of effective stress (1), uw the pore water pressure (kPa), (ua–uw) the matric suction (kPa), fb the angle expressing the rate of increase in strength relative to the matric suction (1). In saturated conditions, the apparent cohesion (the third term on the right-hand side of

Modelling river-bank-erosion processes and mass failure mechanisms

221

equation (9.2) disappears so equation (9.2) reduces to the classical Mohr–Coulomb criterion. 3.1.

Methods of analysis

The application of stability analyses is common in the bank-erosion literature. The analysis of slide failures is typically performed using a Limit Equilibrium Method (LEM) to compute the factor of safety (F), deﬁned as the ratio between stabilising and destabilising forces. Since the 1960s, speciﬁc methods of bank stability analysis have been progressively disseminated, with an increasing effort to deﬁne closed-form solutions for planar failures representative of characteristic bank geometries (Table 9.1). It is evident that research has progressively sought to account for: (1) a more realistic bank geometry and the inﬂuence of tension cracks (Osman and Thorne, 1988); (2) positive pore water pressures and hydrostatic conﬁning pressures (Simon et al., 1991; Darby and Thorne, 1996b); (3) the effects of negative pore water pressures in the unsaturated part of the bank (Rinaldi and Casagli, 1999; Casagli et al., 1999; Simon et al., 2000); and (4) the inﬂuence of riparian vegetation (Abernethy and Rutherfurd, 1998, 2000, 2001; Simon and Collison, 2002; Rutherfurd and Grove, 2004; Pollen et al., 2004; Van de Wiel and Darby, 2004; Pollen and Simon, 2005; Pollen, 2006). Recently, more complex analyses have been utilised for river bank studies (Abernethy and Rutherfurd, 2000; Dapporto et al., 2001, 2003; Simon et al., 2002; Rinaldi et al., 2004) by using various LEM solutions extended to rotational slides (i.e., Bishop, Fellenius, Jambu, Morgestern, GLE) that include features that overcome many of the previous limitations. These analyses provide the following advantages: (1) rotational or composite slide surfaces and generic bank geometries can be deﬁned; (2) either the Mohr–Coulomb or Fredlund et al. (1978) failure criterion can be selected depending on whether the soil conditions are saturated or unsaturated, respectively; (3) a generic pore water pressure distribution can be deﬁned, and conﬁning pressures due to the river can be accounted for; (4) it is possible to perform several analyses for a large number of different sliding surface types and positions, providing more conﬁdence in the identiﬁcation of the most critical failure surface. On the other hand, it is important to recognise that LEM analyses also have some important limitations (Duncan and Wright, 2005). The main one is probably the fact that the mass delimited by the sliding surface is assumed to not be subject to deformation. In other words, only the stresses along the failure surface are accounted for, not the stress distribution within the soil mass. In order to characterise this deformation processes, more complex and sophisticated models used for slope analyses, namely stress-deformation analysis, are required (Grifﬁths and Lane, 1999; Collison, 2001). Such models have not yet been employed speciﬁcally for riverbanks, due to some main reasons: (1) stress-deformation analyses are particularly data-demanding and complex to use; (2) riverbank failures typically occur rapidly, whereas stress-deformation analyses are typically applied to slow landslides, deep-seated deformation, and/or progressive failures on large slopes (e.g., Wiberg et al., 2005; Hu¨rlimann et al., 2006).

222

Table 9.1.

Summary of methods of stability analysis applied to river banks. Mechanism of failure and bank geometry

New capabilities (compared to previous methods)

Main limitations

Typical applications

Main references

‘Culmann’

Planar failure, uniform bank slope

Simple to use

Massive silt or clay, incised rivers of southeastern – midwestern U.S.

Thorne et al. (1981); Thorne (1982)

Thorne & Tovey

Cantilever failure

Composite banks

Osman & Thorne (O&T)

Planar failure with tension crack; bank proﬁle taking into account basal erosion and relic tension crack Planar failure, uniform bank slope

First method speciﬁc for cantilever failure More realistic geometry including effects of basal erosion

Simpliﬁed geometry; failure surface passing from the bank toe; pore water pressures not included Data required not easily available Failure surface passing from the bank toe; pore water pressures not included

Homogeneous, steep, cohesive banks

Thorne and Tovey (1981); Thorne (1982); Van Eerdt (1985) Osman and Thorne (1988); Thorne and Abt (1993)

Simpliﬁed bank geometry (no tension crack)

Homogeneous, steep, cohesive banks

Simon et al.

Failure surface not passing from the bank toe; positive pore pressures and conﬁning pressures incorporated

Simon et al. (1991)

M. Rinaldi, S.E. Darby

Analysis

Planar failure with vertical tension crack, O&T geometry

Rinaldi & Casagli

Planar failure with vertical tension crack, uniform bank slope

Casagli et al.

Planar failure with vertical tension crack, O&T geometry

More realistic geometry

Simon et al.

Planar failure with vertical tension crack, O&T geometry Planar (wedge-type) failure

Layered bank materials

USDA Bank stability model Various commercial software packages

Slides (planar, rotational, composite); generic bank geometry

More realistic geometry with positive pore pressures and conﬁning pressures incorporated Negative pore water pressures taken into account

Incorporates soil reinforcement and surcharge due to vegetation Generic bank geometry and failure surfaces; possible to account for main vegetative mechanical effects

Unsaturated conditions are not considered

Homogeneous, steep, cohesive banks

Darby and Thorne (1996b)

Simpliﬁed bank geometry; simpliﬁed assumptions on water table during drawdown Homogeneous material; Relation river stage – water table needs to be speciﬁed Relation river stage – water table needs to be speciﬁed Simpliﬁed bank geometry

River banks formed in partially saturated soils; rivers with relatively rapid drawdown

Rinaldi and Casagli (1999)

Homogeneous, steep, cohesive river banks formed in partially saturated soils

Casagli et al. (1999), Rinaldi and Casagli (1999)

Layered cohesive river banks formed in partially saturated soils Vegetated river banks

Simon et al. (2000)

Generally more datademanding; requires expertise

When pore water pressure changes at the intra-event scale need to be accounted; rotational or other non-planar failure surfaces and generic bank geometry

Simon and Collison (2002) Dapporto et al. (2001, 2003); Rinaldi et al. (2004)

Modelling river-bank-erosion processes and mass failure mechanisms

Darby & Thorne

223

M. Rinaldi, S.E. Darby

224 3.2.

Effects of pore water pressures

Changes in pore water content and pressures are recognised as one of the most important factors controlling the onset and timing of bank instability (Thorne, 1982; Springer et al., 1985) and the incorporation of these factors in bank process models is one of the major areas of recent progress. Pore water has at least four main effects: (1) reducing shear strength; (2) increasing the unit weight of the bank material; (3) providing an additional destabilising force due to the presence of water in tension cracks (i.e., the force of the water on the sides of the cracks before it inﬁltrates into the soil material); (4) providing additional (stabilising or destabilising) seepage forces. A crucial point when accounting for pore water pressures is their extremely transient character, driven as they are by dynamic hydrological variables (rainfall, river hydrograph). The actual mechanisms and timing of failure induced by pore water pressure effects are difﬁcult to predict if their temporal changes, both at seasonal and intra-event time scales, are not accounted for (Rinaldi and Casagli, 1999; Casagli et al., 1999; Simon et al., 2000). For this reason, bank stability response at the intraevent time scale requires knowledge of the dynamics of saturated and unsaturated seepage ﬂows. Various studies (Dapporto et al., 2001, 2003; Rinaldi et al., 2001, 2004) have made use of the software Seep/w (Geo-Slope International Ltd) to perform two-dimensional, ﬁnite element seepage analyses (Fig. 9.3A) based on the mass conservation equation in a form extended to unsaturated conditions (Fredlund and Rahardjo, 1993): @ @H @ @H @y kx þ ky þQ¼ (9.3) @x @x @y @y @t where H is the total head (m), kx the hydraulic conductivity in the x-direction (m/s), ky the hydraulic conductivity in the y-direction (m/s), Q the unit ﬂux passing in or out of an elementary cube (in this case an elementary square, given that the equation is in two-dimensions) (m2/m2s), y the volumetric water content (m3/m3), t the time (s). Positive and negative pore water pressure distributions obtained by the seepage analysis are then used as input data for the stability analysis; the latter performed using the software Slope/w (Geo-Slope International Ltd.) for application of the LEM. Findings derived from the Rinaldi et al. (2004) analysis have important implications for understanding mass failure processes in relation to the driving hydrologic variables and their dominance in the ﬂuvial system. For example, they partly support Figure 9.3. Seepage and stability analysis of a riverbank of the Sieve River. (Modiﬁed from Rinaldi et al. (2004)). (A) Geometry of the problem, showing ﬁnite element mesh, bank material layers (a, massive silty ﬁne sand; b1, sand; b2, sand with cobbles included; c, silty sand; d, packed sand, gravel, and cobbles; e, loosely packed gravel and cobbles), and their properties (c0 ¼ effective cohesion; f0 ¼ friction angle in terms of effective stress; fb ¼ friction angle in terms of matric suction; g ¼ bulk unit weight; n ¼ porosity; ksat ¼ saturated conductivity; n/a ¼ data not available). (B) Results of the 14/12/1996 ﬂow event: rainfall, river stages, groundwater levels (referred to at a constant distance of 0.5 m from the bank proﬁle), and trend of the safety factor. (C) Minimum safety factor for the simulated ﬂow events as a function of peak river stage: (1) single-peak hydrographs; (2) multiple-peak hydrographs. (Reproduced with permission from Wiley and Sons, 2004.)

Modelling river-bank-erosion processes and mass failure mechanisms

225

M. Rinaldi, S.E. Darby

226

previous authors (Thorne, 1982; Springer et al., 1985; Lawler et al., 1997b) who argued that bank failures occur primarily during the drawdown phase, but they are also better able to discriminate the details of this effect. In particular, it is evident that bank failure in this case often occurs in the very early stage of drawdown (Fig. 9.3B), due to relatively small changes in the motivating and resisting forces. Indeed, it is not necessary for the bank to be saturated to explain bank failure, as would be the case if stability was limited by ‘worst case’ conditions. A second implication is related to the ﬁnding that prolonged and complex hydrographs, with subsidiary peaks preceding the main one, are more destabilising than ﬂow events with a single, distinct, rising phase (Rinaldi et al., 2004; see Fig. 9.3C). Given that the shape of the hydrograph tends to vary systematically with location in the drainage basin, it follows that different destabilising responses can be expected in different locations. In particular, the upper reaches of drainage basins are generally characterised by simple hydrographs with relatively low and distinct peaks, while the ﬂood hydrograph generally tends to become more complex and have a longer duration in downstream reaches. Consequently, pore water pressure distributions may favour the triggering of mass failures in downstream reaches, consistent with the bank process dominance model introduced earlier (Fig. 9.2). It also follows that bank failure frequency and intensity can be promoted by climatic regimes and/or network conﬁgurations that favour multi-peaked rather than single hydrographs. A ﬁnal development highlighted by Rinaldi et al. (2004) is the use of animated graphics as a means of visualising pore water dynamics, enabling the transient effects of these changes to be elucidated more clearly (e.g., ‘Simulation 1’ at http://www.dicea.uniﬁ. it/massimo.rinaldi/private/simulations.htm). Despite these recent advances, further progress is still needed to better simulate pore water pressure changes and their impacts on mass failure. One critical point is the difﬁculty of including in a seepage analysis those banks where the proﬁle is undergoing deformation as a result of ﬂuvial erosion. This is because of the need to continuously adapt the ﬁnite element mesh used to model the problem. A ﬁrst attempt to address this issue has been introduced by Dapporto and Rinaldi (2003), and this is discussed in detail later.

3.3.

Effects of vegetation

The effects of vegetation on river bank processes are many and complex, and most are difﬁcult to quantify. A comprehensive review is beyond the immediate scope of this paper, but given that this ﬁeld is one of the areas in which major recent advances in modelling bank stability have occurred, we provide a brief overview of progress made in quantifying the effects of vegetation on river bank failures. The impacts of vegetation on mass failure can be divided into mechanical and hydrological effects, some of which are positive in terms of their impact on bank stability and some of which are negative. The net change in stability induced by vegetation is, therefore, highly contingent on site-speciﬁc factors, both in terms of the characteristics of the bank (hydrology, shape, sedimentology) and the characteristics

Modelling river-bank-erosion processes and mass failure mechanisms

227

of the vegetation. Considering the mechanical effects of vegetation ﬁrst, the net effect of vegetative surcharge can be either beneﬁcial (increase in normal stress and therefore in the frictional component of soil shear strength) or detrimental (increasing the downslope component of gravitational force), depending on such factors as the position of the tree on the bank, the slope of the shear surface, and the friction angle of the soil (Gray, 1978; Selby, 1982). However, the most important mechanical effect that vegetation has on slope stability is the increase in soil strength induced by the presence of the root system, and considerable progress has recently been made in quantifying this effect (Gray, 1978; Wu et al., 1979; Greenway, 1987; Gray and Barker, 2004; Pollen et al., 2004; Pollen and Simon, 2005; Pollen, 2006). Surcharge and root reinforcement have been recently included in bank stability models (Abernethy and Rutherfurd, 1998, 2000, 2001; Simon and Collison, 2002; Van de Wiel and Darby, 2004; Rutherfurd and Grove, 2004; Pollen and Simon, 2005; Pollen, 2006). In terms of the inﬂuence of riparian vegetation on local-scale river bank hydrology, three main factors can be distinguished: (a) interception; (b) inﬁltration; (c) evapotranspiration. Although these various hydrologic effects are well understood at a conceptual level (e.g., Greenway, 1987; Thorne, 1990), they are in practice extremely difﬁcult to quantify and include in river bank stability models. One exception is the study of Simon and Collison (2002), who quantiﬁed the balance between potential stabilising and destabilising effects based on monitoring data from a river bank along Goodwin Creek, Mississippi (USA). A key ﬁnding of their research is that the hydrologic effects are comparable in magnitude to the mechanical effects of vegetation, and can be either beneﬁcial or detrimental, depending on antecedent rainfall. However, the rate and amount by which plants alter the watercontent distribution within a river bank depend on a great many factors related to vegetation type, soil characteristics, seasonal variations, and climatic conditions of the region. This again makes the effects of vegetation highly contingent and sitedependent, so that generalisation of results from this single study can only be attempted with extreme caution. In addition to the complexity induced by the several and interacting effects of vegetation, a further factor limiting the reliability of prediction can also be mentioned here. Speciﬁcally, the stability of a riverbank is not only dependent on the sitespeciﬁc characteristics of that bank, but it is also conditioned by channel processes operating at the reach scale. Van De Wiel and Darby (2004) have investigated this effect in a series of numerical experiments, demonstrating that reach-scale variations in bed topography induced by the presence of bank vegetation inﬂuences local river bank retreat in a spatially variable manner. The magnitude of this effect was found to be sufﬁciently variable that, in some circumstances, local-scale changes in bank retreat resulting from the presence of vegetation on the bank were less than the changes forced by reach-scale variations in bed topography induced by vegetation assemblages located on the banks in reaches upstream and downstream. These ﬁndings demonstrate that at-a-site analysis by itself is not always sufﬁcient to determine the net beneﬁcial or adverse impact on bank stability of a speciﬁc assemblage of riparian vegetation.

M. Rinaldi, S.E. Darby

228 4.

Concluding discussion: Modelling hydraulic and geotechnical interactions

The preceding sections have identiﬁed how bank retreat involves an interaction between speciﬁc erosion processes and mechanisms. Moreover, in the middle reaches of drainage basins where gravels often dominate, retreat is likely to be driven by a combination of the hydraulic forces of the ﬂow, and mass failures driven by gravity (Fig. 9.2). This is not to exclude the importance of weathering processes, but in this conceptualisation their role is conﬁned to providing a controlling inﬂuence on temporal variations in sediment erodibility (e.g., Prosser et al., 2000, Couper and Maddock, 2001; Lawler, 2005), such that their effect can be accounted for implicitly within ﬂuvial-erosion models. What is clear is that for large extents of gravel-bed reaches in drainage basins, hydraulic and geotechnical factors are both signiﬁcant enough that neither can be ignored (Fig. 9.2). This is not just a question of ensuring that models are comprehensive in the sense that all relevant processes are included. Rather, it is also necessary to capture the interactions between these process groups. This builds on the idea that models that incorporate complex feedbacks, non-linearities, and dynamic interactions between system components are needed to predict behaviour that would otherwise be unforeseen (e.g., Slingerland et al., 1996; Paola, 2000; Bras et al., 2003). In this section we suggest that river bank systems may also behave in this way, by modelling the interactions between hydraulic and geotechnical processes and obtaining predictions with qualitatively different outcomes (in terms of the nature of the onset and timing of bank sediment delivery to the alluvial sedimentary system) than existing models that treat these processes in isolation. While adequate quantitative treatments that include interactions between ﬂuvial erosion and mass failure processes are lacking, basic conceptual models are available. Speciﬁcally, Thorne (1982) has elucidated the concept of basal endpoint control as a framework for understanding the controls on riverbank retreat. The concept is based on the notion that the local bank retreat rate is determined by the status of the sediment budget at the toe of the bank, with Thorne (1982) deﬁning three basal endpoint states as follows.

4.1.

Unimpeded removal

Banks which are in dynamic equilibrium have an approximate balance between the rate at which sediment is supplied to the basal area by ﬂuvial entrainment and mass failure and the removal of this debris by the ﬂow.

4.2.

Excess basal capacity

Here the rate of removal of sediment from the basal region exceeds the rate at which sediment is supplied to the toe, resulting toe erosion may destabilise the bank, increase the rate of retreat, and thus restore a dynamic equilibrium.

Modelling river-bank-erosion processes and mass failure mechanisms 4.3.

229

Impeded removal

Impeded removal is where bank-erosion processes supply material to the base of the bank at a higher rate than it is removed by the ﬂow, such that deposition occurs in the basal zone. Consequently, stability with respect to mass failure increases and the rate of retreat will decrease. The basal endpoint concept is helpful in visualising the coupling that exists between sedimentary processes operating on the banks and those operating in the channel as a whole. Also noteworthy is the point that the residence time of sediment stored at the bank toe is seen as the critical factor controlling long-term bank retreat rates. We return to the signiﬁcance of this below.

4.4. A methodological framework for coupling fluvial erosion, seepage, and bankstability models One of the few attempts to investigate bank-erosion dynamics combining ﬂuvial erosion, pore water pressure changes, and mass bank stability into a single, integrated, modelling approach is the work of Simon et al. (2003), who used three models (Seep/w in combination with the USDA Bank Stability and Toe Erosion (BSM) models) to simulate bank response to ﬂow events. However, although this is undoubtedly a useful exploratory study, it is limited for the following reasons. First, and most signiﬁcantly, the domain of the seepage model is not updated to account for changes in bank geometry caused by ﬂuvial erosion. Instead, the pore waterpressure distributions are calculated for a ﬁxed geometry prior to being imported into the BSM. Consequently, the three modelling components (ﬂuvial erosion, seepage, and mass stability) are not fully coupled, but are instead performed independently. A second limitation of the Simon et al. (2003) study is that they employed a series of artiﬁcial, rectangular-shaped, hydrographs of speciﬁed height and duration in their simulations and it is not clear how these relate to natural ﬂow events observed in the ﬁeld. An alternative example of a numerical simulation of river bank retreat in which ﬂuvial erosion, seepage, and mass failure models are fully integrated is a study of bank dynamics on the Sieve River in Italy (Dapporto and Rinaldi, 2003; Darby et al., 2007). The aim of this simulation was, ﬁrstly, to test the potential of this form of integrated modelling, and secondly to quantify the contribution and mutual role that the various different processes play in controlling bank retreat. What is signiﬁcant is that this research is ﬁrmly grounded in reality (recall that the Sieve River study site had been the focus of earlier bank stability research by Casagli et al., 1999 and Rinaldi et al., 2004). A representative bank proﬁle was used to perform the simulations, using the procedure summarised in Fig. 9.4. For this study we selected a single peak ﬂow event (Q ¼ 792.8 m3/s), that occurred during 18th to 20th November, 1999. For the purposes of the seepage, erosion, and stability analyses, the ﬂow event was discretised into a series of explicit time steps, so that the hydrograph was represented as a succession of steady-state conditions (stepped hydrograph). The time steps were not constant in duration, but were deﬁned according to the variations

M. Rinaldi, S.E. Darby

230

in ﬂow and rainfall, with shorter time steps during phases of rapidly varying ﬂow. A total number of 25 time steps was considered appropriate to represent the ﬂow hydrograph and rainfall inputs in sufﬁcient detail (Fig. 9.5A). As shown in Fig. 9.4, the procedure for modelling riverbank retreat was: (i) to compute the magnitude of ﬂuvial erosion and consequent changes in bank geometry, (ii) to determine the pore water pressure distribution via ﬁnite element seepage analysis, and (iii) to estimate the factor of safety using a slope stability analysis based on the LEM. This sequence is repeated for each subsequent time step, with the bank geometry updated in accordance with any retreat predicted by either of the ﬂuvial erosion or mass failure analyses. Note that while each of the three modelling approaches has already been discussed individually, each requires a particular implementation within the context of the integrated simulation, and these aspects are now discussed.

Initial conditions New time step

near-bank shear stress distribution

τb > τc

Changes in bank geometry

NO

YES LATERAL EROSION Changes in bank profile

Mass failure

FINITE ELEMENT SEEPAGE ANALYSIS Pore water pressure distribution NO

STABILITY ANALYSIS

F<1

YES

Figure 9.4. Flow chart showing the computational logic used for the coupled ﬂuvial erosion– seepage–stability simulation of bank retreat for the 19/11/1999 bankfull event on the Sieve River, Italy.

Modelling river-bank-erosion processes and mass failure mechanisms

231

In order to calculate increments of ﬂuvial erosion in each time step, boundary shear stress was initially estimated using t ¼ gRS, where t ¼ boundary shear stress, g ¼ unit weight of the water, R ¼ hydraulic radius, and S ¼ energy slope. To transform this estimate of the mean boundary shear stress to a bank shear stress value more appropriate for modelling ﬂuvial erosion, a distribution function derived from laboratory ﬂume experiments for rectangular channels (Leutheusser, 1963) was used. This function was applied at each of 32 computational nodes spaced apart up the bank proﬁle at a uniform vertical distance of 0.15 m. Admittedly, this represents a gross simpliﬁcation of the actual near-bank shear stress distribution, but was used in this study merely to demonstrate the methodological approach required to combine the process models. Having obtained the bank stress, the bank proﬁle was deformed by ﬂuvial shear erosion estimated using equation (9.1), with different values of bank erodibility assigned to the different materials within the bank proﬁle (Fig. 9.5B). The ﬁnite element seepage analysis was performed by discretising the river bank proﬁle into about 1600 quadrilateral and triangular elements, and assigning hydraulic conductivity and soil-water characteristic curves to the three layers of bank materials (Fig. 9.5B). In designing ﬁnite-element meshes used in this type of seepage analysis, the discretisation resolution must be veriﬁed to ensure that the output converges to a correct solution, independent of the grid design. The issue of grid veriﬁcation has received attention in the ﬂuid mechanics literature (e.g., Hardy et al., 2003), but to our knowledge no previous river bank study has considered this issue. The discretisation shown in Fig. 9.5B was made with the general criterion of representing the bank area close to the river border in more detail, as this is where the interactions between river stage and pore water pressures are most relevant to mass failure, and it is also where bank retreat is most likely to occur. In contrast, larger cell sizes were employed in areas more distant from the bank–river interface. Regarding the veriﬁcation of this grid, we followed Hardy et al. (2003) by undertaking simulations with a range of cell sizes, conﬁrming that the designed mesh converged to an acceptable solution. The seepage analysis was performed for each time step but, in contrast to previous studies, the bank proﬁle was deformed according to the magnitude of the simulated ﬂuvial erosion. This requires special attention to adapt the ﬁnite element mesh to the new bank geometry at the end of each time step using one of two possible procedures. If the magnitude of the simulated ﬂuvial erosion is less than the width of the boundary cell, the boundary node is shifted inwards by an amount equal to the simulated ﬂuvial erosion. However, as cells cannot be destroyed in the seepage model, this is not possible when the ﬂuvial-erosion increment is greater than the thickness of the boundary cell. Instead, the cell geometry is artiﬁcially modiﬁed by adjusting the hydraulic conductivity in the affected cell(s) to provide a very high transmission rate through the eroded cells (Fig. 9.5C). Subsequently the LEM was used to determine the safety factor for mass failures (for both planar slide and cantilever failure mechanisms) at the end of each time step. Animated graphics (e.g., see ‘Simulation 2’ at http://www.dicea.uniﬁ.it/massimo. rinaldi/private/simulations.htm) provide the best means of visualising the effects of the three interacting processes. However, Fig. 9.5D highlights how ﬂuvial erosion is a quasi-continuous process, active during the entire erosive phase, whereas mass

M. Rinaldi, S.E. Darby

232 99 98 97 96 95 94 93 92 91 18/11/99 18.00

River stage Rainfall

4.E-06 2.E-06 0.E+00 19/11/99 18.00 Time

19/11/99 6.00

20/11/99 6.00 Sediment properties a

b

c

c’(kPa)

2

n/a

n/a

96

φ’(˚)

35

n/a

n/a

95

φ (˚)t

25-35

n/a

n/a

18

n/a

n/a

39.5

35.0

35.0

1.4 10-6 1 10-4

6 10-4

C a

(a) t=0

(b) t=8 h (c) t=8.5 h

b

94

γ(kN/m3) b

93

n (%)

c

ksat (m/s)

92 91

τC(Pa)

90 89 0

Bank retreat (m2)

D

1

2

3

4

6

5

7

8

9

32.5

slide cantilever erosion cohesive erosion non-cohesive hydrograph 3

6

9

12

15

18 21 24 Time (h)

27

30

33

36

39

97.9 97.0 erosion non-cohesive

96.3 95.8

cohesive layer

94.8

non-cohesive layer

erosion cohesive cantilever slide

94.0 93.1 92.3 0.0

0.5

1.0

1.5

n/a

10 11 12 13 14

2.0 1.8 1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0 0

Elevation (m a.s.l.)

1.8

k (m3/Ns-1) 5.4 10-6

2.0 2.5 3.0 Lateral retreat (m)

3.5

4.0

4.5

5.0

99 98 97 96 95 94 93 92 91 90

River stage (m a.s.l.)

Elevation (m a.s.l.)

97

E

8.E-06 6.E-06

98

B

Rainfall intensity (m/s)

1.E-05

River stage (m a.s.l.)

A

27.9 2 10-5

Modelling river-bank-erosion processes and mass failure mechanisms

233

failures exhibit an intermittent, discontinuous behaviour, with frequent cantilever failures occurring both before and during the peak of the event, followed by a slabrotational failure occurring during the recessional phase of the hydrograph. Fig. 9.5E shows how total retreat varies along the bank proﬁle, thereby discriminating dominant process domains. For example, ﬂuvial erosion is the only process responsible for the retreat of the basal granular layer, and although it is still important in the cohesive portion of the bank, mass failures dominate here. The precise mechanism of mass failure also varies with position on the bank, with the slide failure involving most of the cohesive layer, whereas cantilevers prevail on the uppermost portion of the bank. Of particular interest is that these simulation results are qualitatively distinct from conceptual models of bank sediment delivery processes that are founded on eventscale analyses. Previous studies have tended to emphasise mass failure as a quasicatastrophic event, typically timed to occur on the falling limb of event hydrographs. In contrast, our simulation suggests that mass failures occur as a series of erosion episodes, timed at frequent intervals (Fig. 9.6A) as progressive ﬂuvial erosion undermines the bank and trigger failures throughout the ﬂow event. These results can be compared with a simulation of the same ﬂow event in which the bank proﬁle is not deformed by ﬂuvial erosion. For this latter case (Fig. 9.6B) only a single cantilever failure is predicted (at approximately the peak stage of the hydrograph), delivering a total unit (per metre bank length) volume of 0.35 m2 of bank sediment into the channel. This is much less than the total unit volume of 11.65 m2 in the deformingproﬁle scenario (Fig. 9.6A), of which 2.47 m2 emanates from mass failures (slide and cantilevers) and 9.18 m2 from ﬂuvial erosion. Put another way, in the scenario with ﬂuvial erosion, 7 times as much sediment is derived from mass failure and 33 times as much sediment is derived in total compared with the constant geometry case. These differences are not surprising, but they do highlight how effectively ﬂuvial-erosion triggers mass failure. More signiﬁcantly, the effect of ﬂuvial erosion is to induce a quasi-continuous delivery of mass-wasted bank sediment during the early phases, and around the peak, of the ﬂow hydrograph. This is in contrast to models that emphasise mass failure occurring on the falling limb of the hydrograph. An implication is that, even with the much greater volume of bank-derived sediment delivered in the coupled simulation, the residence time of mass-wasted debris delivered to the bank-toe may be much shorter than expected, as the material is injected into the river at times of relatively high ﬂow, rather than on the falling limb. As a result, it seems likely that the basal endpoint status of eroding riverbanks where there is a strong Figure 9.5. Overview of the coupled ﬂuvial erosion–seepage–stability simulation of bank retreat for the 19/11/1999 bankfull event on the Sieve River, Italy. (A) Discretisation of the ﬂow hydrograph; (B) geometry of the problem, showing ﬁnite element mesh, simulated bank material layers (a, massive silty ﬁne sand; b, packed sand, gravel and cobbles; c, loosely packed gravel and cobbles), and their properties (c0 ¼ effective cohesion; f0 ¼ friction angle in terms of effective stress; fb ¼ friction angle in terms of matric suction; g ¼ bulk unit weight; n ¼ porosity; ksat ¼ saturated conductivity; tc ¼ critical shear stress; kd ¼ erodibility coefﬁcient; n/a ¼ data not available); (C) illustration of the procedure used to update the ﬁnite element mesh in the seepage analysis in accordance with simulated ﬂuvial-erosion magnitudes; (D) dynamics of bank-erosion processes simulated during the ﬂow event; (E) total contributions of the different processes to bank retreat as a function of position along the bank proﬁle.

M. Rinaldi, S.E. Darby

234 2,5 2,0

98

Cantilever failure

97

Flow stage

1,5

96

Fluvial erosion

95

1,0

94 0,5

93

0,0

92 0

B

5

10

15

20 Time (h)

25

30

35

40

99

2,5 Sediment Delivered (m3/m)

Flow stage (m a.s.l.)

99 Slide failure

2,0

Cantilever failure

98

Flow stage

97

1,5

96

1,0

95 94

0,5

93

Flow stage (ma.s.l.)

Sediment Delivered (m3/m)

A

92

0,0 0

5

10

15

20 Time (h)

25

30

35

40

Figure 9.6. Comparison of unit sediment volumes predicted to be delivered to the channel by bankerosion processes for the 19/11/1999 event on the Sieve River, Italy. (A) Fully coupled ﬂuvial erosion–seepage–stability analysis; (B) ‘Classical’ bank stability analysis incorporating seepage processes only.

interaction between ﬂuvial erosion and mass failure processes (Fig. 9.2) is quite distinct from riverbanks that can be represented using the ‘classical’ uncoupled approach. This hypothesis requires veriﬁcation, but it might provide a means to explain how some rivers are able to both erode and transmit ﬁne-grained bank sediments effectively enough to produce the very high sediment yields described in the introduction to our review.

Acknowledgements The research reported herein was supported by a Joint Project (Ref: 15077) grant under the Royal Society’s European Science Exchange Programme. Stefano Dapporto is acknowledged for his helpful support and for providing access to the results of his PhD dissertation. We also thank Colin Thorne, Richard Hardy, and two anonymous reviewers for their helpful comments on an earlier version of this paper.

Modelling river-bank-erosion processes and mass failure mechanisms

235

References Abernethy, B., Rutherfurd, I.D., 1998. Where along a river’s length will vegetation most effectively stabilise stream banks? Geomorphology 23, 55–75. Abernethy, B., Rutherfurd, I.D., 2000. The effect of riparian tree roots on the mass-stability of riverbanks. Earth Surf. Process. Landf. 25, 921–937. Abernethy, B., Rutherfurd, I.D., 2001. The distribution and strength of riparian tree roots in relation to riverbank reinforcement. Hydrol. Process. 15, 63–79. American Institute of Aeronautics and Astronautics (AIAA). 1998. Guide for the Veriﬁcation and Validation of Computational Fluid Dynamics Simulations. AIAA Report G-077-1998. American Society of Mechanical Engineers (ASME), 1993. Editorial policy on the control of numerical accuracy. J. Fluids Eng. 115, 339–340. Amiri-Tokaldany, E., Darby, S.E., Tosswell, P., 2003. Bank stability analysis for predicting land loss and sediment yield. J. Am. Water Resour. Assoc. 39, 897–909. Arulanandan, K., Gillogley, E., and Tully, R., 1980. Development of a quantitative method to predict critical shear stress and rate of erosion of natural undisturbed cohesive soils. Report GL-80-5, U.S. Army Engineers, Waterways Experiment Station, Vicksburg, Mississippi. ASCE Task Committee on Hydraulics, Bank Mechanics, and Modeling of River Width Adjustment, 1998. River width adjustment. I: Processes and mechanisms. J. Hydraul. Eng. 124 (9), 881–902. Barker, R., Dixon, L., Hooke, J., 1997. Use of terrestrial photogrammetry for monitoring and measuring bank erosion. Earth Surf. Process. Landf. 22, 1217–1227. Blanckaert, K., Graf, W.H., 2001. Mean ﬂow and turbulence in open-channel bend. J. Hydraul. Eng. 127, 835–847. Bradbrook, K.F., Lane, S.N., Richards, K.S., 2000. Numerical simulation of three-dimensional, timeaveraged ﬂow structure at river channel conﬂuences. Water Resour. Res. 36, 2731–2746. Bras, R.L., Tucker, G.E., Teles, V., 2003. Six myths about mathematical modeling in geomorphology. In: Wilcock, P.R. and Iverson, R.M. (Eds), Prediction in Geomorphology. American Geophysical Union, Washington, DC, pp. 63–79. Bull, L.J., 1997. Magnitude and variation in the contribution of bank erosion to the suspended sediment load of the River Severn, UK. Earth Surf. Process. Landf. 22, 1109–1123. Casagli, N., Rinaldi, M., Gargini, A., Curini, A., 1999. Monitoring of pore water pressure and stability of streambanks: Results from an experimental site on the Sieve River, Italy. Earth Surf. Process. Landf. 24, 1095–1114. Collison, A.J.C., 2001. The cycle of instability: Stress release and ﬁssure ﬂow as controls on gully head retreat. Hydrol. Process. 15, 3–12. Couper, P.R., 2004. Space and time in river bank erosion research: A review. Area 36 (4), 387–403. Couper, P.R., Maddock, I.P., 2001. Subaerial river bank erosion processes and their interaction with other bank erosion mechanisms on the River Arrow, Warwickshire, UK. Earth Surf. Process. Landf. 26, 631–646. Crosato, A., 2007. Effects of smoothing and regridding in numerical meander migration models. Water Resour. Res., 43, W01401, doi:10.1029/2006WR005087. Dapporto, S., 2001. Non-vertical jet testing of cohesive streambank toe material. School of Geography, University of Nottingham, in collaboration with USDA-ARS National Sedimentation Laboratory, Oxford, Mississippi, Internal Report, 41pp. Dapporto, S. and Rinaldi, M., 2003. Modelling of river bank retreat by combining ﬂuvial erosion, seepage and mass failure. Geophys. Res. Abstracts, EGS-AGU-EUG Joint Assembly, Nice, France, 6–11 April 2003. Dapporto, S., Rinaldi, M., Casagli, N., 2001. Mechanisms of failure and pore water pressure conditions: Analysis of a riverbank along the Arno River (Central Italy). Eng. Geol. 61, 221–242. Dapporto, S., Rinaldi, M., Casagli, N., Vannocci, P., 2003. Mechanisms of riverbank failure along the Arno River, Central Italy. Earth Surf. Process. Landf. 28 (12), 1303–1323. Darby, S.E., Rinaldi, M., and Dapporto, S., 2007. Coupled simulations of ﬂuvial erosion and mass wasting for cohesive riverbanks. J. Geophys. Res., 112, doi: 10.1029/2006JF000722.

236

M. Rinaldi, S.E. Darby

Darby, S.E., Spyropoulos, M., Bressloff, N., and Rinaldi, M., 2004. Fluvial bank erosion in meanders: A CFD modelling approach. In: Garcia de Jalon Lastra, D. and Martinez, P.V. (Eds), Aquatic Habitats: Analysis and Restoration, Proceedings of the Fifth International Symposium on Ecohydraulics, September 2004, Madrid, Spain (Volume 1), International Association of Hydraulic Engineering and Research, Madrid, pp. 268–273. Darby, S.E., Thorne, C.R., 1996a. Numerical simulation of widening and bed deformation of straight sand-bed rivers. II: Model evaluation. J. Hydraul. Eng. 122 (4), 194–202. Darby, S.E., Thorne, C.R., 1996b. Development and testing of river-bank stability analysis. J. Hydraul. Eng. 122 (8), 443–454. Duncan, J.M., Wright, S.G., 2005. Soil Strength and Slope StabilityWiley, 295pp. Eaton, B.C., Church, M., Millar, R.G., 2004. Rational regime model of alluvial channel morphology and response. Earth Surf. Process. Landf. 29, 511–529. Fredlund, D.G., Morgenstern, N.R., Widger, R.A., 1978. The shear strength of unsaturated soils. Can. Geotech. J. 15, 312–321. Fredlund, D.G., Rahardjo, H., 1993. Soil Mechanics for Unsaturated Soils. Wiley, Chichester, 482pp. Gerolymos, G.A., Neubauer, J., Sharma, V.C., Vallet, I., 2002. Improved prediction of turbomachinery ﬂows using near-wall Reynolds-stress model. J. Turbomachinery 124, 86–99. Goodson, J.M., Gurnell, A.M., Angold, P.G., Morrissey, I.P., 2002. Riparian seed banks along the lower River Dove UK – their structure and ecological implications. Geomorphology 47, 45–60. Govers, G., 1991. Rill erosion on arable land in central Belgium: Rates, controls and predictability. Catena 18, 133–155. Gray, D.H., 1978. Role of woody vegetation in reinforcing soils and stabilising slopes. Proc. Symp. Soil Reinforcing and Stabilising Techniques, Sydney, Australia, pp. 253–306. Gray, D.H., Barker, D., 2004. Root–soil mechanics and interaction. In: Bennett, S.J. and Simon, A. (Eds), Riparian Vegetation and Fluvial Geomorphology. American Geophysical Union, Washington, DC, pp. 113–124. Greenway, D.R., 1987. Vegetation and slope stability. In: Anderson, M.G. and Richards, K.S. (Eds), Slope Stability. Wiley, pp. 187–230. Grifﬁn, E.R., Kean, J.W., Vincent, K.R., et al., 2005. Modeling effects of bank friction and woody bank vegetation on channel ﬂow and boundary shear stress in the Rio Puerco, New Mexico. J. Geophys. Res., 110, F04023, doi:10.1029/2005JF000322. Grifﬁths, D.V., Lane, P.A., 1999. Slope stability analysis by ﬁnite elements. Geotechnique 49 (3), 387–403. Grissinger, E.H., 1982. Bank erosion of cohesive materials. In: Hey, R.D., Bathurst, J.C., and Thorne, C.R. (Eds), Gravel-bed Rivers. Wiley, Chichester, pp. 273–287. Hanson, G.J., 1990. Surface erodibility of earthen channels at high stresses. Part II – developing an in situ testing device. Trans. Am. Soc. Agric. Eng. 33 (1), 132–137. Hanson, G.J., Simon, A., 2001. Erodibility of cohesive streambeds in the loess area of the midwestern USA. Hydrol. Process. 15, 23–38. Hardy, R.J., Lane, S.N., Ferguson, R.I., Parsons, D.R., 2003. Assessing the credibility of a series of computational ﬂuid dynamic simulations of open channel ﬂow. Hydrol. Process. 17, 1539–1560. Hey, R.D., Bathurst, J.C., Thorne, C.R. (Eds), 1982. Gravel-bed Rivers. Wiley, Chichester, 875pp. Hodskinson, A., Ferguson, R., 1998. Modelling separation ﬂow in meander bends. Hydrol. Process. 12, 1323–1338. Hooke, J.M., 1980. Magnitude and distribution of rates of river bank erosion. Earth Surf. Process. Landf. 5, 143–157. Howard, A.D., 1994. Badlands. In: Abrahams, A.D. and Parsons, A.J. (Eds), Geomorphology of Desert Environments. Chapman and Hall, London, pp. 213–242. Hu¨rlimann, M., Ledesma, A., Corominas, J., Prat, P.C., 2006. The deep-seated slope deformation at Encampadana, Andorra: Representation of morphologic features by numerical modelling. Eng. Geol. 83 (4), 343–357. Kartha, V.C., Leutheusser, H.J., 1972. Distribution of tractive force in open channels. J. Hydraul. Div. ASCE 96, 1469–1483.

Modelling river-bank-erosion processes and mass failure mechanisms

237

Kean, J.W., Smith, J.D., 2004. Flow and boundary shear stress in channels with woody bank vegetation. In: Bennett, S.J. and Simon, A. (Eds), Riparian Vegetation and Fluvial Geomorphology. American Geophysical Union, Washington, DC, pp. 237–252. Kean, J.W. and Smith, J.D., 2006a. Form drag in rivers due to small-scale natural topographic features: 1. Regular sequences. J. Geophys. Res., 111, F04009, doi:10.1029/2006JF000467. Kean, J.W. and Smith, J.D., 2006b. Form drag in rivers due to small-scale natural topographic features: 2. Irregular sequences. J. Geophys. Res., 111, F04010, doi:10.1029/2006JF000490. Knapen, A., Poesen, J., Govers, G., et al., 2007. Resistance of soils to concentrated ﬂow erosion: A review. Earth Sci. Rev. 80, 75–109. Knight, D., Demetriou, J.D., Hamed, M.E., 1984. Boundary shear in smooth rectangular channels. J. Hydraul. Eng. 110, 405–422. Lane, E.W., 1955. Design of stable channels. Trans. Am. Soc. Civil Eng. 120, 1–34. Lane, S.N., Bradbrook, K.F., Richards, K.S., et al., 2000. Secondary circulation cells in river channel conﬂuences: Measurement artefacts or coherent ﬂow structures. Hydrol. Process. 14, 2047–2071. Lane, S.N., Chandler, J.H., Richards, K.S., 1994. Developments in monitoring and terrain modelling small-scale river-bed topography. Earth Surf. Process. Landf. 19, 349–368. Lawler, D.M., 1992. Process dominance in bank erosion systems. In: Carling, P.A. and Petts, G.E. (Eds), Lowland Floodplain Rivers: Geomorphological Perspectives. Wiley, pp. 117–143. Lawler, D.M., 1993. The measurement of river bank erosion and lateral channel change. Earth Surf. Process. Landf. 18, 777–821. Lawler, D.M., 2005. Deﬁning the moment of erosion: The principle of thermal consonance timing. Earth Surf. Process. Landf. 30(13), 1597–1615. Lawler, D.M., Bull., L., Harris, N.M., 1997a. Bank erosion events and processes in the Upper Severn. Hydrol. Earth System Sci. 1, 523–534. Lawler, D.M., Thorne, C.R., Hooke, J.M., 1997b. Bank erosion and instability. In: Thorne, C.R., Hey, R.D., and Newson, M.D. (Eds), Applied Fluvial Geomorphology for River Engineering and Management. Wiley, Chichester, pp. 137–172. Leutheusser, H.J., 1963. Turbulent ﬂow in rectangular ducts. Proc. Am. Soc. Civil Eng. 89, HY3. Lohnes, R., Handy, R.L., 1968. Slope angles in friable loess. J. Geol. 76, 247–258. Millar, R.G., 2000. Inﬂuence of bank vegetation on alluvial Channel patterns. Water Resour. Res. 36 (4), 1109–1118. Millar, R.G., Quick, M.C., 1993. Effect of bank stability on geometry of gravel rivers. J. Hydraul. Eng. 119, 1343–1363. Nagihara, S., Mulligan, K.R., Xiong, W., 2004. Use of a three-dimensional laser scanner to digitally capture the topography of sand dunes in high spatial resolution. Earth Surf. Process. Landf. 29, 391–398. Nicholas, A.P., Sambrook-Smith, G.H., 1999. Numerical simulation of three-dimensional ﬂow hydraulics in a braided channel. Hydrol. Process. 13, 913–929. Nicholas, A.P., Walling, D.E., 1997. Modelling ﬂood hydraulics and overbank deposition on river ﬂoodplains. Earth Surf. Process. Landf. 22, 59–77. Nicholas, A.P., Walling, D.E., 1998. Numerical modelling of ﬂoodplain hydraulics and suspended sediment transport and deposition. Hydrol. Process. 12, 1339–1355. Osman, A.M., Thorne, C.R., 1998. Riverbank stability analysis. Part I: Theory. J. Hydraul Div. ASCE 114 (2), 125–150. Paola, C., 2000. Quantitative models of sedimentary basin ﬁlling. Sedimentology 47, 121–178. Partheniades, E., 1965. Erosion and depostion of cohesive soils. J. Hydraul. Div. ASCE 91, 105–139. Pollen, N., 2006. Temporal and spatial variability of root reinforcement of streambanks: Accounting for soil shear strength and moisture. Catena 69, doi:10.1016/j.catena.2006.05.004. Pollen, N. and Simon, A., 2005. Estimating the mechanical effects of riparian vegetation on streambank stability using a ﬁber bundle model. Water Resour. Res. 41, W07025. doi:10.1029/ 2004WR003801 Pollen, N., Simon, A., Collison, A., 2004. Advances in assessing the mechanical and hydrologic effects of riparian vegetation on streambank stability. In: Bennett, S.J. and Simon, A. (Eds), Riparian Vegetation and Fluvial Geomorphology. American Geophysical Union, Washington, DC, pp. 125–140.

238

M. Rinaldi, S.E. Darby

Prosser, I.P., Hughes, A.O., Rutherfurd, I.D., 2000. Bank erosion of an incised upland channel by subaerial processes: Tasmania, Australia. Earth Surf. Process. Landf. 25, 1085–1101. Reneau, S.L., Drakos, P.G., Katzman, D., et al., 2004. Geomorphic controls on contaminant distribution along an ephemeral stream. Earth Surf. Process. Landf. 29, 1209–1223. Rinaldi, M., Casagli, N., 1999. Stability of streambanks formed in partially saturated soils and effects of negative pore water pressures: The Sieve River (Italy). Geomorphology 26 (4), 253–277. Rinaldi, M., Casagli, N., Dapporto, S., Gargini, A., 2004. Monitoring and modelling of pore water pressure changes and riverbank stability during ﬂow events. Earth Surf. Process. Landf. 29 (2), 237–254. Rinaldi, M., Dapporto, S., Casagli, N., 2001. Monitoring and modelling of unsaturated ﬂow and mechanisms of riverbank failure in gravel bed rivers. In: Nolan, T. and Thorne, C.R. (Eds), Gravel Bed Rivers 2000 CD-ROM. Special Publication of the New Zealand Hydrological Society. Rodi, W., Ferziger, J.H., Breuer, M., Pourquie, M., 1997. Status of large eddy simulation: Results of a workshop. J. Fluids Eng. 119, 248–261. Rutherfurd, I.D., Grove, J.R., 2004. The inﬂuence of trees on stream bank erosion: Evidence from rootplate abutments. In: Bennett, S.J. and Simon, A. (Eds), Riparian Vegetation and Fluvial Geomorphology. American Geophysical Union, Washington, DC, pp. 141–152. Selby, M.J., 1982. Hillslope Materials and Processes. Oxford University Press, Oxford, UK. Simon, A., 1995. Adjustment and recovery of unstable alluvial channels: Identiﬁcation and approaches for engineering management. Earth Surf. Process. Landf. 20, 611–628. Simon, A., Collison, A., 2002. Quantifying the mechanical and hydrologic effects of riparian vegetation on streambank stability. Earth Surf. Process. Landf. 27, 527–546. Simon, A., Curini, A., Darby, S.E., Langendoen, E.J., 2000. Bank and near-bank processes in an incised channel. Geomorphology 35, 193–217. Simon, A., Darby, S.E., 2002. Effectiveness of grade-control structures in reducing erosion along incised river channels: The case of Hotophia Creek, Mississippi. Geomorphology 42, 229–254. Simon, A., Langendoen, E.J., Collison, A., and Layzell, A., 2003. Incorporating bank-toe erosion by hydraulic shear into a bank-stability model: Missouri River, Eastern Montana. Proceedings, EWRL-ASCE, World Water and Environmental Resources Congress, Cd-Rom, 11. Simon, A., Thomas, R.E., Curini, A., Shields, F.D. Jr., 2002. Case study: Channel stability of the Missouri River, Eastern Montana. J. Hydraul. Eng. 128 (10), 880–890. Simon, A., Wolfe, W.J., and Molinas, A., 1991. Mass-wasting algorithms in an alluvial channel model. Proceedings of the 5th Federal Interagency Sedimentation Conference, Las Vegas, Nevada, 2, 8–22 to 8–29. Simons, D.B., Sentu¨rk, F., 1977. Sediment Transport Technology. Water Resources Publications, Fort Collins, CO. Slingerland, R.L., Kump, L.R., Arthur, M.A., Barron, E.J., 1996. Estuarine circulation in the Turonian Western Interior Seaway of North America. Geol. Soc. Am. Bull. 108, 941–952. So, R.M.C., Speziale, C.G., Launder, B.E., 1993. Near-Wall Turbulent Flows. Elsevier, New York. Sotiropoulos, F., 2001. Progress in modelling 3D shear ﬂows using RANS equations and advanced turbulence closures. In: Tzabiras, G.D. (Ed.), Calculation of Complex Turbulent Flows, Advances in Fluid Mechanics Series. WIT Press, Southampton, pp. 209–248. Springer, F.M. Jr., Ullrich, C.R., Hagerty, D.J., 1985. Streambank stability. J. Geotech. Eng. 111 (5), 624–640. Thorne, C.R., 1982. Processes and mechanisms of river bank erosion. In: Hey, R.D., Bathurst, J.C., and Thorne, C.R. (Eds), Gravel-bed Rivers. Wiley, Chichester, pp. 227–271. Thorne, C.R., 1990. Effects of vegetation on riverbank erosion and stability. In: Thornes, J.B. (Ed.), Vegetation and Erosion. Wiley, pp. 125–144. Thorne, C.R., Abt, S.R., 1993. Analysis of riverbank instability due to toe scour and lateral erosion. Earth Surf. Process. Landf. 18, 835–843. Thorne, C.R., Murphey, J.B., and Little, W.C., 1981. Bank stability and bank material properties in the blufﬂine streams of northwest Mississippi. Appendix D. Report to the U.S. Army Corps of Engineers, Vicksburg District Ofﬁce, on Stream Channel Stability, 258pp.

Modelling river-bank-erosion processes and mass failure mechanisms

239

Thorne, C.R., Tovey, N.K., 1981. Stability of composite river banks. Earth Surf. Process. Landf. 6, 469–484. Thorne, S.D., Furbish, D.J., 1995. Inﬂuences of coarse bank roughness on ﬂow within a sharply curved river bend. Geomorphology 12, 241–257. Tolhurst, T.J., Black, K.S., Shayler, S.A., et al., 1999. Measuring the in situ erosion shear strength of intertidal sediments with the Cohesive Strength Meter (CSM). Estuarine, Coastal Shelf Sci. 49, 281–294. Van De Wiel, M.J., Darby, S.E., 2004. Numerical modeling of bed topography and bank erosion along tree-lined meandering rivers. In: Bennett, S.J. and Simon, A. (Eds), Riparian Vegetation and Fluvial Geomorphology. American Geophysical Union, Washington, DC, pp. 267–282. Van Eerdt, M.M., 1985. Salt marsh cliff stability in the Oosterschelde. Earth Surf. Process. Landf. 10, 95–106. Walling, D.E., Owens, P.N., Leeks, G.J.L., 1999. Fingerprinting suspended sediment sources in the catchment of the River Ouse, Yorkshire, UK. Hydrol. Process. 13, 955–975. Wiberg, N.E., Koponen, M., Runesson, K., 2005. Finite element analysis of progressive failure in long slopes. Int. J. Numer. Anal. Met. 14 (9), 599–612. Wu, T.H., McKinnell, W.P., Swanston, D.N., 1979. Strength of tree roots and landslides on Prince of Wales Island, Alaska. Can. Geotech. J. 16 (1), 19–33.

Discussion by Gary Williams The bank-erosion model has been based on an actual river bank. Presumably actual bank retreat was measured for a given rainfall and ﬂood hydrograph. How was the erosion model calibrated against actual bank retreat? What was adjusted to ﬁt, and why was that adjustment approach used?

Reply by the authors In the erosion model, the erodibility coefﬁcient of the basal packed gravel is the calibration parameter, being deﬁned by forcing best agreement between calculated and measured ﬁnal bank toe retreat at the end of the simulation. This is done because reliable methods for measuring erodibility parameters for the packed gravel are not available.

Modelling river-bank-erosion processes and mass failure mechanisms

239

Thorne, C.R., Tovey, N.K., 1981. Stability of composite river banks. Earth Surf. Process. Landf. 6, 469–484. Thorne, S.D., Furbish, D.J., 1995. Inﬂuences of coarse bank roughness on ﬂow within a sharply curved river bend. Geomorphology 12, 241–257. Tolhurst, T.J., Black, K.S., Shayler, S.A., et al., 1999. Measuring the in situ erosion shear strength of intertidal sediments with the Cohesive Strength Meter (CSM). Estuarine, Coastal Shelf Sci. 49, 281–294. Van De Wiel, M.J., Darby, S.E., 2004. Numerical modeling of bed topography and bank erosion along tree-lined meandering rivers. In: Bennett, S.J. and Simon, A. (Eds), Riparian Vegetation and Fluvial Geomorphology. American Geophysical Union, Washington, DC, pp. 267–282. Van Eerdt, M.M., 1985. Salt marsh cliff stability in the Oosterschelde. Earth Surf. Process. Landf. 10, 95–106. Walling, D.E., Owens, P.N., Leeks, G.J.L., 1999. Fingerprinting suspended sediment sources in the catchment of the River Ouse, Yorkshire, UK. Hydrol. Process. 13, 955–975. Wiberg, N.E., Koponen, M., Runesson, K., 2005. Finite element analysis of progressive failure in long slopes. Int. J. Numer. Anal. Met. 14 (9), 599–612. Wu, T.H., McKinnell, W.P., Swanston, D.N., 1979. Strength of tree roots and landslides on Prince of Wales Island, Alaska. Can. Geotech. J. 16 (1), 19–33.

Discussion by Gary Williams The bank-erosion model has been based on an actual river bank. Presumably actual bank retreat was measured for a given rainfall and ﬂood hydrograph. How was the erosion model calibrated against actual bank retreat? What was adjusted to ﬁt, and why was that adjustment approach used?

Reply by the authors In the erosion model, the erodibility coefﬁcient of the basal packed gravel is the calibration parameter, being deﬁned by forcing best agreement between calculated and measured ﬁnal bank toe retreat at the end of the simulation. This is done because reliable methods for measuring erodibility parameters for the packed gravel are not available.

Modelling river-bank-erosion processes and mass failure mechanisms

239

Thorne, C.R., Tovey, N.K., 1981. Stability of composite river banks. Earth Surf. Process. Landf. 6, 469–484. Thorne, S.D., Furbish, D.J., 1995. Inﬂuences of coarse bank roughness on ﬂow within a sharply curved river bend. Geomorphology 12, 241–257. Tolhurst, T.J., Black, K.S., Shayler, S.A., et al., 1999. Measuring the in situ erosion shear strength of intertidal sediments with the Cohesive Strength Meter (CSM). Estuarine, Coastal Shelf Sci. 49, 281–294. Van De Wiel, M.J., Darby, S.E., 2004. Numerical modeling of bed topography and bank erosion along tree-lined meandering rivers. In: Bennett, S.J. and Simon, A. (Eds), Riparian Vegetation and Fluvial Geomorphology. American Geophysical Union, Washington, DC, pp. 267–282. Van Eerdt, M.M., 1985. Salt marsh cliff stability in the Oosterschelde. Earth Surf. Process. Landf. 10, 95–106. Walling, D.E., Owens, P.N., Leeks, G.J.L., 1999. Fingerprinting suspended sediment sources in the catchment of the River Ouse, Yorkshire, UK. Hydrol. Process. 13, 955–975. Wiberg, N.E., Koponen, M., Runesson, K., 2005. Finite element analysis of progressive failure in long slopes. Int. J. Numer. Anal. Met. 14 (9), 599–612. Wu, T.H., McKinnell, W.P., Swanston, D.N., 1979. Strength of tree roots and landslides on Prince of Wales Island, Alaska. Can. Geotech. J. 16 (1), 19–33.

Discussion by Gary Williams The bank-erosion model has been based on an actual river bank. Presumably actual bank retreat was measured for a given rainfall and ﬂood hydrograph. How was the erosion model calibrated against actual bank retreat? What was adjusted to ﬁt, and why was that adjustment approach used?

Reply by the authors In the erosion model, the erodibility coefﬁcient of the basal packed gravel is the calibration parameter, being deﬁned by forcing best agreement between calculated and measured ﬁnal bank toe retreat at the end of the simulation. This is done because reliable methods for measuring erodibility parameters for the packed gravel are not available.

Gravel-Bed Rivers VI: From Process Understanding to River Restoration H. Habersack, H. Pie´gay, M. Rinaldi, Editors r 2008 Elsevier B.V. All rights reserved.

241

10 Adjustment of the bed surface size distribution of gravel-bed rivers in response to cycled hydrographs Gary Parker, Marwan A. Hassan and Peter Wilcock

Abstract Mountain gravel-bed rivers typically display a surface layer that is armored. That is, the surface layer visible at low ﬂow is coarser than both the substrate and mean annual bedload transported. The surface layer is difﬁcult to sample at the high ﬂows that transport most of the gravel. As a result, the question as to whether the surface layer remains armored at high ﬂows is something of a mystery. The few measurements available suggest that some form of armoring may be in place at high ﬂows as well. In lieu of more measurements, numerical modelling provides an avenue to explore this issue. Research results are presented using a 1D model of aggradation and degradation to mobile-bed equilibrium in gravel-bed streams. In the model, a hydrograph is cycled repeatedly so that water discharge goes up and down in time. The magnitude of the bedload feed rate and the size distribution of the feed material are, however, held constant at the upstream end of the reach. As a result, the ﬁnal mobile-bed equilibrium attained is characterized by a bed at the upstream end of the reach that cyclically degrades and coarsens at high ﬂow (when the sediment feed rate is not sufﬁcient) and aggrades and becomes ﬁner at low ﬂow (when there is an excess of sediment feed). Only a short distance downstream, however, a remarkable tradeoff occurs. The bed adjusts so that over the great majority of the modelled reach the bed elevation and surface size distribution become invariant in time, hardly changing at all from low ﬂow to high ﬂow. The bedload transport rate and size distribution, however, ﬂuctuate strongly with the hydrograph. That is, the higher ﬂows support a higher transport rate of coarser material and the lower ﬂows support a lower transport rate of ﬁner material. The implication is that rivers subject to repeated hydrographs can evolve so that neither surface grain size distribution nor mean bed elevation (averaged over bars) need change much with ﬂow, nearly all the variation

E-mail addresses: [email protected] (G. Parker), [email protected] (M.A. Hassan), [email protected] (P. Wilcock) ISSN: 0928-2025

DOI: 10.1016/S0928-2025(07)11127-5

G. Parker, M.A. Hassan, P. Wilcock

242

being absorbed by the bedload. If this is true, it provides a most useful result; the surface grain size distribution seen at low ﬂow may be very close to that seen at high ﬂow. The results have been veriﬁed with two transport relations, that of Parker and that of Wilcock and Crowe. The reasons behind this simple result are explored in terms of a ‘‘hydrograph boundary layer,’’ downstream of which the effect of the hydrograph on bed elevation and surface size distribution become negligible. The results of the numerical model also indicate that for a given hydrograph, the degree to which the surface is armored relative to the grain size distribution of the feed sediment decreases with increasing gravel feed rate. 1.

Introduction

The topic of this paper can be introduced in terms of a metaphor. The metaphor is based on a three-panel cartoon from the ‘‘Far Side’’ series of Gary Larson (1984). In the top panel, four cows are standing on their two hind legs, looking perfectly at ease, when one of them shouts ‘‘car!’’. Just before the car passes by, all four cows hurriedly change to a four-legged stance. After the car passes by, the cows relax and revert to standing on their two hind legs. The implication is that only when we see them, the cows are standing on four legs, but when we cannot see them, they are standing on two legs. There is no way for us to verify that they really stand on two legs when we are not looking, because the act of our looking causes them to change their stance to four legs. So we have to accept their two-legged stance as an article of faith. Many rivers show a coarse surface layer, or armor at low ﬂow. An example is given in Fig. 10.1 in terms of the River Wharfe, U.K. It has often been assumed that this armor is ‘‘washed out,’’ or at least strongly subdued, during ﬂood ﬂows. But it is precisely in the middle of ﬂood ﬂows when it is impossible to verify this hypothesis. For example in Fig. 10.2a the Elbow River, Canada is shown at low ﬂow; in Fig. 10.2b it is shown at a ﬂow estimated to be close to the 100-year ﬂood. The Elbow River is known to be armored at low ﬂow (Hollingshead, 1971). Is it armored at high ﬂow, or not? Is there any way of ﬁnding this out? Some evidence suggests that the armor layer might still be present in at least some form at ﬂood ﬂows capable of moving most of the available sizes. For example, Parker et al. (1982) and Parker and Klingeman (1982) report experiments on gravel transport for which an armor layer was observed to be in place even under conditions of rather intense bedload transport. Andrews and Erman (1986) performed measurements in Sagehen Creek, USA during a snowmelt storm which mobilized gravel sizes coarser than the surface median size, and found that an armor layer similar to the one observed at low ﬂow remained in place. More recently Wilcock et al. (2001) performed experiments on mobile-bed transport of heterogeneous gravel in which the armor layer was not only present for all ﬂows, but varied little over a wide range of ﬂow conditions. More recently, Wilcock and DeTemple (2005) have provided an indirect indication of the persistence of an armor layer over a ﬂood in Oak Creek, USA. They used the surface-based transport relation of Wilcock and Crowe (2003) and the measured data for bedload magnitude and size distribution of Milhous (1973) to back-calculate the surface grain size distribution. Their results indicate a

Adjustment of the bed surface size distribution of gravel-bed rivers

243

Figure 10.1. View of the armored bed of the River Wharfe, U.K. Image courtesy D.M. Powell.

surface grain size distribution that does not change as ﬂow and transport rate increase, instead remaining essentially the same as the surface grain size distribution measured at low ﬂow. So is it possible to say that gravel-bed rivers should always be armored? The answer, of course, is no. Fig. 10.3 shows the Nahal Yatir in Israel, a gravel-bed river which is unarmored at low ﬂow, and might be assumed to be unarmored at high ﬂow (Powell et al., 2001). Such streams are relatively common in arid environments. This paper is focused on the way in which a gravel-bed river adjusts to a cycled hydrograph that ranges from low ﬂow to ﬂood ﬂow. The analysis is based on a 1D numerical model. It yields a fascinating conclusion: rivers can adjust over cycled hydrographs so that their surface layers become invariant to the speciﬁc discharge of the hydrograph. The conclusion is partial and tentative, because real rivers are not 1D entities and numerical models are less than perfect expressions of reality. If the result is borne out by more detailed ﬁeld and experimental measurements, however, it

244

G. Parker, M.A. Hassan, P. Wilcock

Figure 10.2. (a) View of the Elbow River, Canada at low ﬂow. Image courtesy A.B. Hollingshead. (b) View of the same reach of the Elbow River, Canada at a ﬂood estimated to be a 100-year recurrence ﬂood. Image courtesy A.B. Hollingshead.

has the potential to greatly simply gravel transport calculations. In particular, it implies that the surface grain size distribution to be used in a surface-based calculation of bedload transport during a ﬂood may be approximated by the surface distribution measured at low ﬂow (when the bed is accessible). Or put in terms of the introductory metaphor, the cows are likely always standing on four legs, whether we are watching or not.

Adjustment of the bed surface size distribution of gravel-bed rivers

245

Figure 10.3. View of the unarmored bed of the Nahal Yatir, Israel. Image courtesy D.M. Powell.

2.

Summary of the conﬁguration and the essential result

The very simple conﬁguration illustrated in Fig. 10.4 is considered. The river is treated as a sediment-feed ﬂume with constant width, vertical sidewalls and speciﬁed length. Downstream bed elevation is held constant. Water is fed into the ﬂume at a rate that varies cyclically, so representing a repeated hydrograph including a wide range of discharges. In implementing the hydrograph, it is possible to discard ﬂows at which bed sediment is not moved in significant amounts. A smooth hydrograph must be discretized into steps in order for implementation in a numerical model. Gravel (gravel/sand) is also fed into the ﬂume at the same upstream point as the water. The software developed for the present study allows for the gravel input rate to vary cyclically in time as well. In the present implementations, however, the rate and grain size distribution of the upstream gravel (gravel/sand) are held constant. If the above conﬁguration is sustained for a sufﬁciently long time, a mobile-bed equilibrium is eventually reached. The equilibrium can strictly be deﬁned, however, only as an average over the hydrograph. What are the characteristics of this equilibrium? Since the input rate and grain size distribution of the feed sediment are held constant over the repeated hydrograph, one might expect that the bed cyclically degrades and coarsens during the higher ﬂows of the hydrograph, and then aggrades and becomes ﬁner over the lower ﬂows of the hydrograph. The numerical results reported here suggest, however, that this behavior is limited to a very short reach downstream of the feed point. This reach may be termed a ‘‘hydrograph boundary

G. Parker, M.A. Hassan, P. Wilcock

246 qbTf , pbfi initial bed profile La

mobile-bed equilibrium profile (averaged over hydrograph) Qw

η

L Figure 10.4. tures.

Schematic diagram of the conﬁguration for the numerical experiments with sediment mix-

layer,’’ where ‘‘boundary layer’’ is used in the mathematical sense of a restricted zone over which a variable changes strongly (Nayfeh, 1993). Downstream of this hydrograph boundary layer a remarkable tradeoff occurs. Bed slope and bed surface grain size distribution become essentially invariant with discharge on the hydrograph (and thus time) and distance downstream. Instead, the magnitude and grain size distribution of the bedload vary cyclically with position on the hydrograph (but do not vary with distance downstream).

3.

Quantiﬁcation of the conﬁguration

Again the very simple conﬁguration illustrated in Fig. 10.4 is considered. The river has constant width B, no ﬂoodplain, vertical sidewalls and constant length L. Sidewall effects are neglected for simplicity. The downstream bed elevation Z is ﬁxed at 0; thus where x denotes streamwise distance from the sediment feed point, Zjx¼L ¼ 0

(10.1)

Water is introduced to the ﬂume at the upstream end (x ¼ 0) at ﬂow discharge Qw(t), where t denotes time. Here Qw(t) is allowed to vary cyclically so as to simulate a periodic hydrograph. The river is assumed to be sufﬁciently steep and the reach sufﬁciently short so that ﬂood waves traverse the reach in a time that is very short compared to the characteristic time of morphodynamic evolution. As a result, all ﬂows can be accurately described using the normal (steady, uniform) approximation, with the discharge at any point in the ﬂow (nearly) instantaneously adjusting to the upstream value Qw(t). This assumption is often reasonable for reaches of mountain gravel-bed rivers not longer than a few 10’s of km. A heterogeneous mixture of gravel (or gravel and sand) that moves as bedload is also fed in at the upstream end of the ﬂume. For simplicity, this is assumed to be fed in at a constant rate, and to have a constant feed grain size distribution. Consider N grain size bins i ¼ 1..N, each with characteristic size Di. Let qbi denote the volume bedload transport rate per unit width in the ith grain size range. The

Adjustment of the bed surface size distribution of gravel-bed rivers

247

total volume bedload transport rate per unit width summed over all grain sizes is given as qbT ¼

N X

qbi

(10.2)

i¼1

The volume fraction of bedload in the ith grain size range is denoted as q pbi ¼ bi qbT

(10.3)

The feed values of qbi, qbT and pbi are denoted respectively as qbfi, qbfT and pfi, such that qbi jx¼0 ¼ qbfi

(10.4a)

qbT jx¼0 ¼ qbTf

(10.4b)

pbi jx¼0 ¼ pbfi

(10.4c)

Here these feed values are held constant over the hydrograph. That is, in a given numerical experiment the magnitude and grain size distribution of the sediment feed are held constant, even though the water discharge is varying cyclically.

4.

Exner equation of sediment conservation

The Exner equation of sediment conservation for mixtures used here is that of Parker and Sutherland (1990). The bed is divided into an upper ‘‘active’’ (‘‘surface,’’ ‘‘exchange’’) layer and a lower substrate. The active layer consists of the gravel that exchanges directly with bedload transport. The active layer exchanges sediment with the substrate as the bed aggrades (transfer from active layer to substrate) or degrades (transfer from substrate to active layer). As noted in Fig. 10.4, the active layer has thickness La(x, t), the notation indicating that thickness may vary in both time and distance downstream. The volume fraction of material in the ith grain size range in the active layer, or surface fraction Fi may also vary in x and t, but the active layer is approximated as having no vertical structure. As the bed aggrades or degrades, the fraction in the ith grain size range exchanged at the interface between the bottom of the surface layer and the top of the substrate is denoted as fIi. Denoting bed porosity as lp, the grain size-speciﬁc Exner equation of sediment continuity takes the form @ @ @q (10.4) ð1 lp Þ f Ii ðZ La Þ þ ðF i La Þ ¼ bi @x @t @t Summing the above equation over all grain sizes and recalling that N X i¼1

Fi ¼

N X i¼1

f Ii ¼ 1

(10.5a,b)

G. Parker, M.A. Hassan, P. Wilcock

248 it is found that

@Z @q ¼ bT @x @t Reducing (10.4) with (10.6) results in the relation @F i @La @q @q ð1 lp Þ La þ ðF i f Ii Þ ¼ f Ii bT bi @t @t @x @x ð1 lp Þ

(10.6)

(10.7)

Thus (10.6) describes the evolution of bed elevation and (10.7) describes the evolution of the surface grain size distribution. In implementing the above equation it is necessary to specify the thickness La of the active layer and the interfacial exchange fractions fIi. Here the thickness of the active layer is speciﬁed as an order-one multiple of the surface size Ds90 such that 90 percent of the sediment is ﬁner: La ¼ na Ds90

(10.8)

where na is an order-one dimensionless parameter. The interfacial fractions are speciﬁed as follows. The substrate is likely to contain its own stratigraphy, so that substrate fractions fi vary with vertical distance z (in addition to x). As the bed degrades, the substrate just below it is mined into the active layer. As the bed aggrades, some mixture of bedload and active layer material is transferred to the substrate. Thus 8 @Z < f i z¼ZL ; @t o0 a (10.9) f Ii ¼ @Z : aF i þ ð1 aÞpbi ; @t 40 where a is a speciﬁed parameter between 0 and 1 (Hoey and Ferguson, 1994; ToroEscobar et al., 1996).

5.

Flow hydraulics

The ﬂow hydraulics is computed using a simple normal (steady, uniform) ﬂow approximation that is often suitable for mountain streams. Water discharge Qw is related to water discharge per unit width qw, ﬂow depth H and depth-averaged ﬂow velocity U as Qw ¼ qw B ¼ UHB

(10.10)

The Manning–Strickler relation of Parker (1990) is used to compute ﬂow resistance; 1=6 U H 1=2 ¼ Cf ¼ ar (10.11) un ks In the above relation Cf is a dimensionless bed friction coefﬁcient, ar takes a value of 8.1, ks denotes a roughness height, related here to surface size Ds90 as ks ¼ nk Ds90

(10.12)

Adjustment of the bed surface size distribution of gravel-bed rivers

249

where nk is another order-one coefﬁcient here set equal to 2 and un denotes a shear velocity, related to bed shear stress tb as tb ¼ ru2n

(10.13)

where r denotes water density. In the present analysis all resistance is assumed to be skin friction; form drag is neglected for simplicity. According to the normal ﬂow approximation, tb is related to the depth-slope product as tb ¼ rgHS

(10.14)

where S denotes bed slope. Between equations (10.10), (10.11), (10.13) and (10.14) it is found that the shear velocity un can be computed from the local bed slope S and local roughness height ks (and the water discharge per unit width qw, which varies in time but not space) as un ðx; tÞ ¼ a3=10 g7=20 ½qw ðtÞ3=10 ½ks ðx; tÞ1=20 Sðx; tÞ7=20 r

(10.15)

Note that the above equation is dimensionally homogeneous.

6.

Surface-based bedload transport formulation

The analysis presented here is restricted to the case of bedload transport of either gravel, or gravel with some admixture of sand. In order to compute the evolution of the fractions Fi in the surface (active, exchange) layer as the bed evolves, it is necessary to tie the bedload transport rate of the ith size range to the availability of this size range in the surface layer. Several formulae are presently available to do this, including Parker (1990), Powell et al. (2001), Hunziker and Jaeggi (2002) and Wilcock and Crowe (2003). Here calculations are performed with the relation of Wilcock and Crowe (2003). It should be pointed out, however, that calculations with the relation of Parker (1990) illustrate that the essential conclusions of the analysis are independent of the speciﬁc bedload formulation used. All of the above bedload transport relations can be cast in a form such that they predict a dimensionless bedload transport rate W ni for the ith grain size range, which is related to the volume transport rate per unit width qbi as qbi ¼ F i

u3n W Rg i

(10.16)

where R denotes the submerged speciﬁc gravity of the sediment, given as R¼

rs 1 r

(10.17)

and rs denotes the material density of the sediment. For natural sediments R is often close to 1.65. Details of the relation of Wilcock and Crowe (2003) are not presented here; these can be found in the original reference and Parker (2004). Instead, brief summaries of input parameters are given.

G. Parker, M.A. Hassan, P. Wilcock

250

In the case of the relation of Wilcock and Crowe (2003), in order to compute W ni it is necessary to know (a) the shear velocity un , (b) the submerged speciﬁc gravity of the sediment R, (c) the surface grain sizes and fractions (Di, Fi) and (d) the surface geometric mean size Dsg and the fraction of sand Fs in the surface layer, both of which can be computed from (Di, Fi). Note that sand is specifically included, and that varied sand content in the surface layer can have a strong effect on the transport rate of gravel-sized material.

7.

Flow of the calculation

In order to perform a calculation, the following dimensionless parameters must be speciﬁed in advance: bed porosity lp, coefﬁcient na in (10.8) describing active layer thickness, coefﬁcient a describing transfer to the substrate as the bed aggrades, coefﬁcient ar in the resistance relation (10.11), coefﬁcient nk in the relation for the roughness height (10.12) and the submerged speciﬁc gravity R of the sediment. In addition, the characteristic grain sizes Di, i ¼ 1yN must be speciﬁed. Finally, the cyclic variation of water discharge per unit width with time qw(t) must be speciﬁed. The ﬂow of the calculation is as follows. At any given time t the bed proﬁle Z(x, t), and surface fractions Fi(x, t) are taken to be known. Bed slope S is computed as @Z (10.18) @x The surface fractions Fi are used to compute Ds90, Dsg, Fs everywhere. La and ks are everywhere computed from (10.8) and (10.12), respectively. Shear velocity u ðx; tÞ is then computed everywhere from (10.15). A knowledge of un , Ds90, Dsg, Fs allows computation of W ni from the relation of Wilcock and Crowe (2003), and thus qbi(x, t) everywhere from (10.16). The parameters qbT and pbi are then computed from (10.2) and (10.3). The bed elevation proﬁle one time step later, i.e., at time t+Dt, is then computed from a discretized version of (10.6), i.e., S¼

Zðx; t þ DtÞ ¼ Zðx; tÞ

1 @qbT Dt ð1 lp Þ @x

Equation (10.6) also directly estimates @Z/@t; @Z 1 @qbT ¼ @t t 1 lp @x t

(10.19)

(10.20)

which then allows evaluation of fIi everywhere from (10.9). The surface fractions one time step later are then evaluated from (10.7) as 1 @q @q ðF i f Ii Þ @La Dt (10.21) f Ii bT bi F a jtþDt ¼ F a jt þ @x @x @t ð1 lp ÞLa La In principle, (10.21) requires an iterative solution due to the presence of the term @La/@t, but this term is typically small, and can usually be evaluated from the previous time step.

Adjustment of the bed surface size distribution of gravel-bed rivers

251

The boundary conditions on the above formulation are (10.1), which speciﬁes a ﬁxed downstream bed elevation, and (10.4a) and (10.4b), which specify a ﬁxed feed rate and feed grain size distribution of sediment. The initial conditions consist of a speciﬁed initial bed slope, and speciﬁed initial grain size distributions for the bed surface and substrate. The reach of length L is discretized into M intervals, each with length Dx ¼ L/M, bounded by M+1 nodes. The node k ¼ 1 denotes the node farthest upstream and the node k ¼ M+1 denotes the node farthest downstream. Sediment is fed in at a ghost node one step upstream of the node k ¼ 1. Spatial derivatives involving sediment transport parameters (qbT, qbi) are computed using an upwinding scheme, e.g., at the kth node, qbT;k qbT;k1 qbT;kþ1 qbT;k @qbT þ ð1 au Þ (10.22) ¼ au @x k Dx Dx In the above relation au is an upwinding coefﬁcient. A value of au of 0.5 corresponds to a central difference scheme, and a value of au satisfying the conditions 0.5oaur1 corresponds to an upwinded scheme.

8.

Outline of and input for the numerical runs

Numerical runs are performed for both a cycled hydrograph and a constant ﬂow corresponding to the average of that hydrograph. The hydrograph chosen for implementation is a discretization of a 4.5-day symmetrical triangular hydrograph with a beginning and end ﬂow discharge per unit width of 2 m2/s and a maximum ﬂow discharge per unit width of 20 m2/s. The average ﬂow of the hydrograph is 10 m2/s. The hydrograph and its average ﬂow are shown in Fig. 10.5. The hydrograph (or its equivalent constant ﬂow) is run once per year for 4.5 days. The river is taken to be morphologically inactive for the rest of the year. The sediment feed is taken to be the bimodal mix of gravel and sand given Fig. 10.6. This distribution has median size Dl50f (l ¼ load, 50 ¼ median, f ¼ feed) of 32 mm, a geometric mean size Dlgf (l ¼ load, g ¼ geometric mean, f ¼ feed) of 16.22 mm and a fraction sand Fslf (s ¼ sand, l ¼ load, f ¼ feed) of 0.25. The initial surface and substrate size distributions at t ¼ 0 are taken to be identical to that of the feed sediment. The following parameters are speciﬁed in the calculation as follows: R ¼ 1.65, na ¼ 2, nk ¼ 2, lp ¼ 0.4, ar ¼ 8.1, a ¼ 0.5, au ¼ 0.75. The initial bed slope is a constant speciﬁed value. This value was chosen to be not too far from the expected value in order to minimize the amount of computational time to reach a ﬁnal mobile-bed equilibrium. In all numerical experiments L ¼ 20,000 m and M ¼ 20, so that Dx ¼ 1000 m. Eleven runs, i.e., Runs 1 H, 2 H, 3Hy 11H were conducted with the hydrograph of Fig. 10.5. Eleven more runs, i.e., 1C, 2C, 3Cy 11C were conducted with the average ﬂow of the hydrograph also shown in Fig. 10.5. The input parameters are speciﬁed in Table 10.1 (hydrograph runs) and Table 10.2 (constant-ﬂow runs). It is seen there that sediment feed rates qbTf vary over a wide range, from 1 106 m2/s to

G. Parker, M.A. Hassan, P. Wilcock

252

Water discharge per unit width qw, m2/s

25

20 hydrograph

15

10 constant flow with average of hydrograph

5

0 0

1

2

3

4

5

days

Figure 10.5. Plot of the 4.5-day hydrograph used in the runs of Table 10.1, and the 4.5-day period of constant ﬂow with the average water discharge per unit width of the hydrograph used in the runs of Table 10.2. Each 4.5-day ﬂow (hydrograph or constant ﬂow) was repeated once per year.

Feed Sediment, Percent Finer

100 90 80 70 60 50 40 30 20 10 0 0.1

1

10

100

1000

D (mm) Figure 10.6. Grain size distribution of the feed sediment. The same grain size distribution was used for the initial surface and substrate size distributions.

1 101 m2/s. Also shown in the Table 10.1 are initial bed slope SI, time step during a ﬂood Dtf, number of time steps per step on the ﬂow hydrograph (during the ﬂood) nstep and time duration Tdur of the calculation. Also shown in Table 10.2 are SI, Tdur and time step Dt (real time including ﬂood time and inactive time).

Adjustment of the bed surface size distribution of gravel-bed rivers

253

Table 10.1.

Input parameters for the runs with a repeated hydrograph.

Run

qbTf (m2/s)

SI

Dtf (days)

nstep

Tdur (years)

1H 2H 3H 4H 5H 6H 7H 8H 9H 10H 11H

1 101 3.5 102 1 102 3.5 103 1 103 3.5 104 1.5 104 1 104 3.5 105 1 105 1 106

0.0263263 0.0124 0.00571 0.00335 0.00208 0.00157 0.00133 0.00125 0.00112 0.00102 0.000976

0.00625 0.0125 0.025 0.025 0.1 0.25 0.5 0.5 0.5 0.5 0.5

80 40 20 20 5 2 1 1 1 1 1

300 600 1200 3000 3000 6000 12000 12000 24000 36000 120000

Table 10.2.

Input parameters for the runs with an equivalent constant ﬂow.

Run

qbTf (m2/s)

SI

Dt (days)

Tdur (years)

1C 2C 3C 4C 5C 6C 7C 8C 9C 10C 11C

1 101 3.5 102 1 102 3.5 103 1 103 3.5 104 1.5 104 1 104 3.5 105 1 105 1 106

0.0263 0.0124 0.00571 0.00335 0.00208 0.00157 0.00133 0.00125 0.00190 0.00170 0.00150

0.457 1.826 7.305 14.61 58.44 116.88 116.88 116.88 58.44 58.44 58.44

30 120 480 960 3840 15360 30720 30720 30720 30720 61440

9.

Results for constant ﬂow at mobile-bed equilibrium

Before analyzing the case of a hydrograph, it is of value to brieﬂy discuss the results for the equivalent constant ﬂows. The results analyzed here are for the 11 runs of Table 10.2. Recall that water discharge per unit width is held at 10 m2/s for 4.5 days of the year for all runs, but that the sediment feed qbTf rate varies from a high rate of 1 101 m2/s for Run 1C to a low of 1 106 m2/s for Run 11C. Fig. 10.7 shows plots of the following three parameters versus sediment feed rate qbTf at ﬁnal, mobile-bed equilibrium; bed slope S, surface geometric mean size Dsg and geometric mean size of the feed sediment Dlgf. Note that the feed rate qbTf is everywhere equal to the transport rate qbT in the case of mobile-bed equilibrium at constant ﬂow. The diagram shows that bed slope S increases, and surface geometric mean size Dsg decreases with increasing qbTf. At very low values of qbTf the bed approaches a static

G. Parker, M.A. Hassan, P. Wilcock

254 static armor 100

mobile armor

unarmored 0.03

0.025

0.02

10

feed geom. mean

0.015

S

Dsg (mm), Dlgf (mm), S

surface geom. mean

Dlgf Dsg S

0.01

S 1 1.E-06

1.E-05

1.E-04

1.E-03

1.E-02

0.005

0 1.E-01

qbTf (m2/s) Figure 10.7. Plot of surface geometric mean size Dsg and bed slope S as functions of the volume bedload feed rate per unit width qbTf for the runs of Table 10.2, i.e., constant ﬂow. The parameters pertain to mobile-bed equilibrium. Also shown is the geometric mean size Dlgf of the feed sediment, which is held constant in all runs.

armor, and S and Dsg show values near 0.0013 and 100 mm, respectively, that change only weakly with qbTf. S increases ever more sharply with higher values of qbTf, reaching a value near 0.026 for qbTf ¼ 0.1 m2/s. Likewise, the surface geometric mean size Dsg decreases toward the geometric mean of the feed sediment at high value of qbTf, indicating that the surface layer is approaching an unarmored state compared to the sediment transported. In between these two limits is a wide range for which (a) the bed is armored under mobile-bed conditions, and (b) the geometric mean size of the armor gradually becomes ﬁner with increasing qbTf. This result suggests that a mobile-bed armor might be expected under ﬂood conditions prevailing in a stream such as the River Wharfe (Fig. 10.1), which likely has modest gravel supply, but might be absent under ﬂood conditions in a stream such as the Nahal Yatir (Fig. 10.3), which has an extremely high gravel supply (Powell et al., 2001). In Fig. 10.8 the fraction of sand in the surface layer Fss (ﬁrst s ¼ sand, second s ¼ surface) and the fraction of sand in the bedload feed material feed Fslf (s ¼ sand, l ¼ load, f ¼ feed) are plotted against qbTf. (The plot shows percentages rather than fractions.) The percentage sand in the feed Fslf has been held constant at 25 percent for all runs. At the lowest feed rate qbTf of 1 106 m2/s the surface layer contains only 0.56 percent sand. The percentage of sand in the surface layer rises with increasing qbTf, until at the highest feed rate qbTf a value of 19 percent is attained. Thus sand is nearly absent from the surface layer at sediment transport rates low enough to correspond to a nearly static armor, even though the load is 25 percent sand. When the sediment transport is sufﬁciently high, the percent sand in the surface layer

Adjustment of the bed surface size distribution of gravel-bed rivers

255

100

Fslf, Fss (both x 100)

percent sand in feed

10 Fslf Fss percent sand in surface 1

0.1 1.E-06

1.E-05

1.E-04

1.E-03

1.E-02

1.E-01

2

qbTf (m /s) Figure 10.8. Plot of the fraction sand Fss in the surface layer as a function of volume bedload feed rate per unit width qbTf for the runs of Table 10.2, i.e., constant ﬂow. The parameters pertain to mobile-bed equilibrium. Also shown is the fraction sand Fslf in the feed sediment.

approaches that of the load (nearly unarmored). In between is a range of mobile armor for which sand is present in the surface layer, but at a noticeably lower percentage than in the load. Fig. 10.9 shows the grain size distributions for the feed, and for the surface layer at mobile-bed equilibrium for all 11 runs of Table 10.2. The progression with increasing feed rate qbTf from nearly static armor to mobile armor, and then to a state for which the bed is nearly unarmored is readily apparent from the ﬁgure.

10. Results for cycled hydrographs: formation and significance of the hydrograph boundary layer The 11 runs with cycled hydrographs are those summarized in Table 10.1. Before discussing the results of the numerical modelling, however, it is important to review the constraints on the experiments. In every numerical run an identical hydrograph (that of Fig. 10.5) is repeated once annually for hundreds to hundreds of thousands of years until a mobile-bed equilibrium state is reached. In each numerical run the sediment feed rate qbTf is held constant over the hydrograph. This constant feed rate varied from 1 101 m2/s in Run 1H to 1 106 m2/s in Run 11H. Because of this conﬁguration, a mobile-bed equilibrium cannot consist of a constant state. Instead, it must consist of a state in which exactly the same cycle is repeated over and over. Now what might this cycling equilibrium consist of ?

G. Parker, M.A. Hassan, P. Wilcock

256 100 90

constant discharge

80

feed qbtF = 1.0e-1m2/s qbTf = 3.5e-2 qbTf = 1.0e-2 qbTf = 3.5e-3 qbTf = 1.0e-3 qbTf = 3.5e-4 qbTf = 1.5e-4 qbTf = 1.0e-4 qbTf = 3.0e-5 qbTf = 1.0e-5 qbTf = 1.0e-6

feed

Percent Finer

70 60 mobile armor

50 40

nearly unarmored

30 20 10 nearly static 0 0.1

1

10 D (mm)

100

1000

Figure 10.9. Surface grain size distributions at mobile-bed equilibrium for all the constant-ﬂow runs of Table 10.2. The bedload feed rates qbTf are speciﬁed in m2/s. Also shown is the grain size distribution of the feed sediment, which is held constant in all runs.

At the very upstream end, water discharge ﬂuctuates up and down, but the sediment feed rate and grain size distribution are held constant. As a result, one might expect the bed to cyclically (a) degrade and coarsen at the high ﬂows, when the transport capacity exceeds the feed rate, and (b) aggrade and become ﬁner at the low ﬂows, when the transport capacity is less than the feed rate. The numerical runs reveal, however, a fascinating result. The above, ‘‘expected’’ behavior is realized only in a relatively short hydrograph boundary layer downstream of the feed point. Downstream of this hydrograph boundary layer a tradeoff takes place. The mobile-bed equilibrium consists of a bed which no longer cycles even though ﬂow discharge continues to cycle. That is, bed elevation and surface grain size distribution remain constant in time over the hydrograph. Instead, the cycling is transferred to the bedload transport rate and bedload grain size distribution. The bedload transport rate cyclically increases, and the bedload becomes coarser at the high ﬂows of the hydrograph, and the pattern is reversed at the low ﬂows of the hydrograph. This pattern is illustrated schematically in Fig. 10.10. In that ﬁgure, bed elevation is shown to vary cyclically only in a short hydrograph boundary layer near the feed point. Within this boundary layer bed elevation Z and surface geometric mean size Dsg ﬂuctuate over the hydrograph in response to changing water discharge qw but constant sediment feed rate qbTf and constant feed grain size distribution (e.g., constant feed geometric mean size Dlgf, where l ¼ load, g ¼ geometric mean, f ¼ feed). Downstream of this boundary layer, over a region that consists of the great majority of the total modelled reach, bed elevation Z and surface geometric mean size

Adjustment of the bed surface size distribution of gravel-bed rivers

257

water surface high flow

qbTf, pbfi

"boundary layer" water surface low flow

qw Dsg

qbT, Dlg

η

t

qbT, Dlg qw

η

Dsg

η

L t

Figure 10.10. Diagram illustrating the essential results of the numerical runs with a cycled hydrograph of Table 10.1. At the upstream end of the reach the total volume feed rate per unit width qbT and the geometric mean size of the load Dlg are held constant, but ﬂow discharge per unit width qw is allowed to vary cyclically. In a short hydrograph boundary layer downstream, bed elevation Z and surface geometric mean grain Dsg vary cyclically as well. Downstream of this boundary layer Z and Dsg become invariant with time, and qbT and Dlg now vary cyclically in time with the ﬂow hydrograph. The diagram pertains to mobile-bed equilibrium.

Dsg remain constant over the hydrograph, but the bedload transport rate qbT and size distribution (e.g., load geometric mean size Dlg) ﬂuctuate cyclically over the hydrograph. Numerical results for two numerical runs, Run 3H (qbTf ¼ 1 102 m2/s) and Run 6H (qbTf ¼ 3.5 104 m2/s) are sufﬁciently characteristic to warrant their use in justifying the above conclusions. Run 3H is considered ﬁrst. Fig. 10.11 shows a plot of the streamwise variation of bed slope at the maximum (peak) ﬂow and the minimum (end) ﬂow of the last hydrograph of the experiment, well after mobile-bed equilibrium had been reached. Bed slope S is identical at the peak and end ﬂows of the hydrograph at all points except those within about 4000 m of the feed point. In this hydrograph boundary layer reach, bed slope is low at the peak ﬂow and high at the low ﬂow, as the bed responds to a ﬂuctuating ﬂow discharge but a constant feed rate. No such adjustment is observed downstream of this short boundary layer reach. Not only does bed slope S become invariant downstream of the hydrograph boundary layer reach, but also surface geometric mean size Dsg no longer varies with the hydrograph. This is illustrated in Fig. 10.12, which shows the grain size distributions of the surface layer at both the maximum (peak) and minimum (end) ﬂow of the last hydrograph of the experiment. The two surface size distributions overlap each other so closely that they are virtually identical. The results for Fig. 10.12 pertain to the node at the end of the model reach (k ¼ 21), where x ¼ L. The same invariance in the surface grain size distribution is also found farther upstream, as long as the point in question is downstream of the hydrograph boundary layer of Fig. 10.11. Fig. 10.12 illustrates the tradeoff between ﬂuctuations in surface size distribution and load size distribution downstream of the hydrograph boundary layer reach. In addition to surface size distributions, the plot shows the size distribution of the load at the maximum (peak) ﬂow and minimum (end) ﬂow of the last hydrograph of the experiment, the load distribution averaged over the last hydrograph and ﬁnally the

G. Parker, M.A. Hassan, P. Wilcock

258 0.0065 qbTf = 1x10-2 m2/s

Slope max flow Slope end flow

Bed Slope S

0.006

0.0055 boundary layer where hydrograph affects bed slope and surface size distribution

0.005 0

5000

10000 Distance m

15000

20000

Figure 10.11. Bed slope proﬁles for the maximum (peak) ﬂow and the minimum (end) ﬂow of the last hydrograph of Run 3H. Mobile-bed equilibrium has been achieved by this time. Bed slope ﬂuctuates with the hydrograph only in a short region (hydrograph boundary layer) near the feed point (x ¼ 0).

100 90

qbTf = 1x10-2 m2/s

80

Percent finer

70

load end

60 50

feed ds final surface max flow ds final surface end flow ds final load max flow ds final load end flow ds final load hydro ave

feed, load averaged, load peak

40 30 20 10 0 0.1

surface peak and end

1

10 Grain size mm

100

1000

Figure 10.12. Grain size distributions of the surface material and bedload at the maximum (peak) and minimum (end) ﬂows of the last hydrograph of Run 3H. Also included are the grain size distributions of the feed sediment and the bedload averaged over the hydrograph. The node in question is the one farthest downstream (x ¼ L). Note that the surface size distribution is nearly invariant, whereas the bedload size distribution varies strongly between the peak and end ﬂows.

Adjustment of the bed surface size distribution of gravel-bed rivers

259

size distribution of the feed sediment. The node in question is again the farthest node downstream. The grain size distribution of the load averaged over the hydrograph is very close to that of the feed, as is to be expected at mobile-bed equilibrium. The load size distribution at the maximum (peak) ﬂow is somewhat coarser than the feed sediment, and the load size distribution at the minimum (end) ﬂow is markedly ﬁner. That is, the grain size distribution of the load now ﬂuctuates markedly with the hydrograph, even as the grain size distribution of the surface layer remains invariant. Again, the same result is obtained farther upstream at any point downstream of the hydrograph boundary layer reach. Fig. 10.13a shows the variation in water discharge per unit width qw, volume sediment transport rate per unit width qbT and bed slope S over the last hydrograph of the run. Again, the plot pertains to the node farthest downstream, and again essentially the same results are obtained at all points downstream of the hydrograph boundary layer. Note that bed slope S remains nearly perfectly constant over the hydrograph. The sediment transport rate qbT, however, ﬂuctuates strongly in concordance with the hydrograph; high ﬂows cause high bedload rates, and low ﬂow causes low bedload transport rates. Fig. 10.13b has the same format as Fig. 10.13a, but the point in question is the ﬁrst node upstream (k ¼ 1; x ¼ 0), just downstream of the ghost node where sediment is fed in. This node is well within the hydrograph boundary layer reach, as is seen from Fig. 10.11; bed slope S ﬂuctuates cyclically as the channel tries to adjust to a constant sediment supply but cyclic water discharge variation. Also plotted in Fig. 10.13b is the variation of load qbT over the hydrograph. Note that qbT is already varying cyclically over the hydrograph at the ﬁrst node upstream, even though the pattern is rather strongly skewed as compared to that in Fig. 10.13a. Fig. 10.14a shows the variation of qw, load geometric mean size Dlg and surface geometric mean size Dsg at the farthest node downstream (M ¼ 21) over the last hydrograph of the numerical run. The corresponding plot for the farthest node upstream (M ¼ 1) is given in Fig. 10.14b. Comparing the two ﬁgures, it is seen that surface geometric mean Dsg grain size varies cyclically at the upstream node, but shows little variation at the downstream node. The load geometric mean grain size varies cyclically at both nodes, but the variation is stronger at the downstream node. The essential points are worth summarizing again. A hydrograph is cycled repeatedly. Sediment is fed in at the upstream end at a constant rate and with a constant grain size distribution. Within a short hydrograph boundary layer near the feed, the mobile-bed equilibrium associated with these constraints consists of a bed elevation, bed slope and surface size distribution that ﬂuctuate cyclically with the hydrograph. Farther downstream, however, bed elevation, bed slope and surface size distribution evolve to become independent of the hydrograph, and instead the cyclic variation is transferred to the magnitude and grain size distribution of the bedload transport. The above summary also implies the conditions under which the hydrograph boundary layer disappears. Suppose that the run were to be continued at mobile-bed equilibrium, but the bedload feed rate and size distribution were now allowed to ﬂuctuate cyclically and sympathetically with the hydrograph, in precisely the way that is observed downstream of the hydrograph boundary layer. Under such

G. Parker, M.A. Hassan, P. Wilcock

260 30

0.1 -2

2

qbTf = 1x10 m /s

qw (m2/s), qbT (m2/s), S

qbT 20

0.01 qw m^2/s qbT m^2/s S

S 10

0.001

qw 0

0.0001 0

1

2

3

4

5

Days

(a) 30

0.1

qw (m2/s), qbT (m2/s), S

qbTf = 1x10-2 m2/s

qbT 20

0.01 qw m^2/s qbT m^2/s S

S 10

0.001

qw 0

0.0001 0 (b)

1

2

3

4

5

Days

Figure 10.13. (a) Plot of the variation of water discharge per unit width qw, bed slope S and volume bedload transport rate per unit width qbT with time over the last hydrograph, at the node farthest downstream (x ¼ L), for Run 3H. Note that slope S is invariant over the hydrograph, but the time variation in the bedload transport rate qbT tracks that of the water discharge per unit width qw. (b) Plot of the variation of water discharge per unit width qw, bed slope S and volume bedload transport rate per unit width qbT with time over the last hydrograph, at the node farthest upstream (x ¼ 0), for Run 3H. Note that slope S and the bedload transport rate qbT both vary in time over the hydrograph.

Adjustment of the bed surface size distribution of gravel-bed rivers

261

50 qbTf = 1x10-2 m2/s

qw (m2/s), Dlg, Dsg (mm)

40 Dsg 30 qw m^2/s Dlg mm Dsg mm

20 Dlg 10 qw 0 0

1

2

(a)

3

4

5

Days

60 qbTf = 1x10-2 m2/s

qw (m2/s), Dlg, Dsg (mm)

50

40

Dsg qw m^2/s Dlg mm Dsg mm

30

20

Dlg

10

qw

0 0 (b)

1

2

3

4

5

Days

Figure 10.14. (a) Plot of the variation of water discharge per unit width qw, load geometric mean size Dlg and surface geometric mean size Dsg with time over the last hydrograph, at the node farthest downstream (x ¼ L), for Run 3H. Note that load geometric mean size Dlg varies strongly in time over the hydrograph, whereas surface geometric mean size Dsg is nearly invariant. (b) Plot of the variation of water discharge per unit width qw, load geometric mean size Dlg and surface geometric mean size Dsg with time over the last hydrograph, at the node farthest upstream (x ¼ 0), for Run 3H. Note that both surface geometric mean size Dsg and load geometric mean size Dlg vary notably over the hydrograph.

G. Parker, M.A. Hassan, P. Wilcock

262

conditions the hydrograph boundary layer would disappear, and the bed elevation and surface size distribution would everywhere become constant over the hydrograph, even inside what used to be the hydrograph boundary layer. The above statement is easily conﬁrmed with the numerical model. All 11 numerical runs indicate that, other factors being equal, the length of the hydrograph boundary layer increases with increasing sediment feed rate qbTf. With this in mind, results are also shown for Run 6H (qbTf ¼ 3.5 104 m2/s). Fig. 10.15 for Run 6H shows the slope proﬁles at the maximum (peak) and minimum (end) ﬂows of the last hydrograph of the run. Again, the node is the one farthest downstream. The hydrograph boundary layer is again clearly apparent, although the phasing of aggradation is different from that seen in Fig. 10.11 for Run 3H. Fig. 10.16 for Run 6H, which corresponds to Fig. 10.12 for Run 3H, again shows that the invariance of the surface grain size distributions over the hydrograph at the last node, even while the grain size distribution of the load varies strongly between the maximum (peak) and minimum (end) ﬂows of the last hydrograph. Figs. 10.17a and 10.17b for Run 6H, which correspond to Fig. 10.13a for Run 3H, show that at the node farthest downstream the bedload transport rate varies strongly 0.0016

qbTf = 3.5x10-4 m2/s

Bed Slope S

0.00159

Slope max flow

0.00158

Slope end flow

0.00157

0.00156

boundary layer where hydrograph affects bed slope and surface size distribution

0.00155

0

5000

10000

15000

20000

Distance m Figure 10.15. Bed slope proﬁles for the maximum (peak) ﬂow and the minimum (end) ﬂow of the last hydrograph of Run 6H. Mobile-bed equilibrium has been achieved by this time. Bed slope ﬂuctuates with the hydrograph only in a short region near the feed point (x ¼ 0). The hydrograph boundary layer is somewhat shorter than in the case of Fig. 10.11 (Run 3H). The very small streamwise variation in bed slope near x ¼ L is an artifact of the model.

Adjustment of the bed surface size distribution of gravel-bed rivers

263

100 90

qbTf = 3.5x10-4 m2/s

80

Percent finer

70

load end feed ds final surface max flow ds final surface end flow ds final load max flow ds final load end flow ds final load hydro ave

60 50 feed, load averaged, load peak

40 30 20

surface peak and end

10 0 0.1

1

10 Grain size mm

100

1000

Figure 10.16. Grain size distributions of the surface material and bedload at the maximum (peak) and minimum (end) ﬂows of the last hydrograph of Run 6H. Also included are the grain size distributions of the feed sediment and the bedload averaged over the hydrograph. The node in question is the one farthest downstream (x ¼ L). Note that the surface size distribution is invariant, whereas the bedload size distribution varies strongly between the peak and end ﬂow.

over the hydrograph at mobile-bed equilibrium, whereas bed slope remains constant. Figs. 10.17c and 10.17d for Run 6H, which correspond to Fig. 10.13b for Run 3H, show that both the bedload transport rate and the bed slope vary over the hydrograph at the node farthest upstream. A comparison of Figs. 10.13b and 10.17d show that a phase shift in bed slope variation at x ¼ 0 of about half a day in Run 6H as compared to Run 3H, such that bed slope attains its maximum and minimum values at later times in Run 6H than in Run 3H. Figs. 10.18a and 10.18b for Run 6H, which corresponding to Figs. 10.14a and 10.14b for Run 3H, illustrate a surface geometric mean size that remains nearly perfectly constant over the hydrograph at the downstream node, but which shows marked cyclic variation at the upstream node. Summarizing, Run 6H shows all the features of Run 3H, but more strongly so. For example, it is of value to compare Fig. 10.18a of Run 6H with Fig. 10.14a of Run 3H. In both cases it is seen that the surface geometric mean size Dsg at the node farthest downstream shows little cyclic variation at mobile-bed equilibrium. Some variation is, however detectable in the case of Fig. 10.14a for Run 3H, where the variation is undetectable in the case of Fig. 10.18a for Run 6H. This reﬂects the tendency for the effect of the upstream boundary conditions to become ever more strongly concentrated in an ever-thinner hydrograph boundary layer reach as qbTf declines.

30

0.01

30

0.002

qw m^2/s qbT m^2/s

q

0.001 20

q 0.0001

10 0.00001

qw (m2/s), qbT (m2/s), S

= 3.5x10 m /s

qw m^2/s S

= 3.5x10 m /s

0.0018

264

qw (m2/s), qbT (m2/s), S

q

S

20

0.0016

0.0014 10 0.0012 q

q

0 1

2

3

4

5

1

2

3

= 3.5x10 m /s

10 0.00001 q

0

0.000001 2

3

Days

4

5

qw m^2/s S

= 3.5x10 m2/s

0.0018 20

S 0.0016

0.0014 10 q

0.0012

0

0.001 0

(d)

1

2

3

4

5

Days

Figure 10.17. (a) Plot of the variation of water discharge per unit width qw and volume bedload transport rate per unit width qbT with time over the last hydrograph, at the node farthest downstream (x ¼ L), for Run 6H. The time variation in the bedload transport rate qbT tracks that of the water discharge per unit width qw. (b) Plot of the variation of water discharge per unit width qw and bed slope S with time over the last hydrograph, at the node farthest downstream (x ¼ L), for Run 6H. Note that slope S is invariant over the hydrograph. (c) Plot of the variation of water discharge per unit width qw and volume bedload transport rate per unit width qbT with time over the last hydrograph, at the node farthest upstream (x ¼ 0), for Run 6H. The time variation in the bedload transport rate qbT tracks that of the water discharge per unit width qw. (d) Plot of the variation of water discharge per unit width qw and bed slope S with time over the last hydrograph, at the node farthest upstream (x ¼ 0), for Run 6H. Note that slope S varies over the hydrograph.

G. Parker, M.A. Hassan, P. Wilcock

0.0001

qw (m2/s), qbT (m2/s), S

q

1

0.002 q

20

5

30

qw m^2/s qbT m^2/s

0.001

4

Days

(b)

0.01 q

0 (c)

0.001 0

Days

30

qw (m2/s), qbT (m2/s), S

0

0.000001 0 (a)

Adjustment of the bed surface size distribution of gravel-bed rivers

265

60

qw (m2/s), Dlg, Dsg (mm)

50

Dsg

qbTf = 3.5x10-4 m2/s

40 qw m^2/s Dlg mm Dsg mm

30

20

Dlg

10

qw

0 0 (a)

1

3

2

4

5

Days

60

qw (m2/s), Dlg, Dsg (mm)

50

Dsg

qbTf = 3.5x10-4 m2/s

40

qw m^2/s Dlg mm Dsg mm

30 Dlg

20

10

0

qw

0 (b)

1

2

3

4

5

Days

Figure 10.18. (a) Plot of the variation of water discharge per unit width qw, load geometric mean size Dlg and surface geometric mean size Dsg with time over the last hydrograph, at the node farthest downstream (x ¼ L), for Run 6H. Note that load geometric mean size Dlg varies strongly in time over the hydrograph, whereas surface geometric mean size Dsg is invariant. (b) Plot of the variation of water discharge per unit width qw, load geometric mean size Dlg and surface geometric mean size Dsg with time over the last hydrograph, at the node farthest upstream (x ¼ 0), for Run 6H. Note that both surface geometric mean size Dsg and load geometric mean size Dlg vary notably over the hydrograph.

11.

Comparison of results for constant ﬂow versus cycled hydrograph

In the ﬁgures discussed below the morphodynamic behavior of the river outside the hydrograph boundary layer at mobile-bed equilibrium is characterized by the

G. Parker, M.A. Hassan, P. Wilcock

266 static armor

mobile armor

unarmored

0.03

0.025

surface geom. mean averaged over hydrograph feed geom. mean

0.02

0.015

10

load geom. mean averaged over hydrograph

S

Dsg (mm), Dlg, Dlgf (mm), S

100

Dlgf Dsga Dlga S

0.01

0.005 S 1 1.E-06

1.E-05

1.E-04

1.E-03

1.E-02

0 1.E-01

2

qbTf (m /s) Figure 10.19. Plot of surface geometric mean size averaged over the hydrograph Dsga, bedload geometric mean size averaged over the hydrograph Dlga and bed slope S as functions of the volume bedload feed rate per unit width qbTf for the hydrograph runs of Table 10.1. The parameters pertain to mobile-bed equilibrium at the node farthest downstream (x ¼ L). Also shown is the geometric mean size Dlgf of the feed sediment.

behavior at the node farthest downstream. This is because at mobile-bed equilibrium the behavior outside the hydrograph boundary layer is, to a high degree of approximation, everywhere the same as that at the node farthest downstream. Fig. 10.19 shows plots of the following parameters at mobile-bed equilibrium against sediment feed rate qbTf for the 11 runs with a cycled hydrograph: surface geometric mean size averaged over the hydrograph Dsga, bedload geometric mean size averaged over the hydrograph Dlga, geometric mean size of the feed sediment Dlsg and bed slope S. All parameters except Dlsg pertain to the node farthest downstream. The overall pattern is identical to that seen in Fig. 10.7 for the runs with constant ﬂow; bed slope S increases, and surface geometric mean size Dsg decreases, with increasing sediment feed rate qbTf. Again, the progression from nearly static armor, to mobile armor, and ﬁnally to a condition for which the bed is nearly unarmored is readily evident. Fig. 10.20 shows the surface sand contents at the maximum ﬂow (peak ﬂow) fraction of sand content Fssp and minimum ﬂow (end ﬂow) fraction of sand content Fsse in the surface layer (expressed in percentages) as functions of qbTf. Again, the data are for the node farthest downstream and the last hydrograph of the run. Note that Fsse and Fssp are essentially identical except at the highest two sediment feed rates. At these high feed rates the hydrograph boundary layer becomes somewhat diffuse, so that some effects of the upstream boundary conditions propagate weakly to the downstream end of the ﬂume. The general pattern of Fig. 10.20, according to

Adjustment of the bed surface size distribution of gravel-bed rivers

267

100 Fslf Fssp Fsse

Fslf, Fssp, Fsse (all x 100)

percent sand in feed

10

percent sand in surface, peak and end flows

1 1.E-06

1.E-05

1.E-04

1.E-03

1.E-02

1.E-01

2

qbTf (m /s)

Figure 10.20. Plot of the fraction sand in the surface Fssp at the maximum (peak) ﬂow and fraction sand in the surface Fsse at the minimum (end) ﬂow as functions of volume bedload feed rate per unit width qbTf for the runs of Table 10.1, i.e., cycled hydrograph. The parameters pertain to mobile-bed equilibrium at the node farthest downstream (x ¼ L). Also shown is the fraction sand Fslf in the feed sediment. 0.03

100

surf. geom. mean feed geom. mean

0.025

hydrograph percent sand

0.02

10

0.015

hydrograph S

constant

0.01

constant

S

Dsg (mm, const), Dsga (mm, hyd). Fss x 100 (const), Fssp x 100 (hydro)

constant

Dsg, const Dsga, hyd Fss const Fssp, hyd Dlgf S, const S, hyd

0.005

hydrograph 1 1.E-06

1.E-05

1.E-04

1.E-03

1.E-02

0 1.E-01

2

qbTf m /s Figure 10.21. Plots of the following parameters at mobile-bed equilibrium as functions of volume bedload feed rate per unit width qbTf: bed slope S for both the hydrograph runs of Table 10.1 and constant-ﬂow runs of Table 10.2; fraction sand in the surface layer Fss for the runs of Table 10.2; fraction sand in the surface layer at the maximum (peak) ﬂow Fssp for the runs of Table 10.1; surface geometric mean size Dsg for the runs of Table 10.2; surface geometric mean size averaged over the hydrograph Dsga for the runs of Table 10.1. Also shown for reference is the geometric mean size of the feed sediment Dlgf.

G. Parker, M.A. Hassan, P. Wilcock

268

which the fraction of sand in the surface layer increases with increasing sediment transport rate, closely parallels that seen in Fig. 10.8 for the runs with constant ﬂow. Fig. 10.21 shows comparisons of the equilibrium values of the following parameters for both the constant-ﬂow runs and the hydrograph runs: (a) surface geometric mean size Dsg for constant ﬂows and Dsgp at the maximum (peak) ﬂows for the hydrograph, (b) fraction of sand in the surface layer Fss for constant ﬂows and Fssp at the maximum (peak) ﬂows for the hydrograph and (c) bed slope S for both cases. Also shown for reference is the geometric mean size Dlgf of the feed sediment. It should be noted that the data for the constant-ﬂow runs pertain to the node farthest upstream, whereas the data for the hydrograph runs pertain to the node farthest downstream. Having said this, in the case of mobile-bed equilibrium at constant discharge all parameters at the downstream node should take the same values as at the upstream node. Fig. 10.21 highlights the effect of the cycled hydrograph as opposed to constant ﬂow corresponding to the average of the hydrograph continued for the same number of days per year. In all cases a cycled hydrograph leads to a bed with a lower slope and a ﬁner surface layer than in the case of the corresponding constant ﬂow. The difference between the two becomes progressively weaker as feed rate qbTf increases. The overall trends are, however, the same in both cases. Fig. 10.22 for the case of cycled hydrographs corresponds to Fig. 10.9 for constant ﬂows. It shows the downstream surface grain size distributions at mobile-bed equilibrium for all the numerical runs of Table 10.1. Also shown is the grain size

100 90 80

feed feed qbtF = 1.0e-1 qbTf = 3.5e-2 qbTf = 1.0e-2 qbTf = 3.5e-3 qbTf = 1.0e-3 qbTf = 3.5e-4 qbTf = 1.5e-4 qbTf = 1.0e-4 qbTf = 3.0e-5 qbTf = 1.0e-5 qbTf = 1.0e-6

repeated hydrographs

Percent Finer

70 60

mobile armor

50 40

nearly unarmored

30 20 10 0 0.1

nearly static armor 1

10

100

1000

D (mm) Figure 10.22. Surface grain size distributions at mobile-bed equilibrium for all the hydrograph runs of Table 10.1. The bedload feed rates qbTf are speciﬁed in m2/s. Also shown is the grain size distribution of the feed sediment. The data pertain to the node farthest downstream (x ¼ L).

Adjustment of the bed surface size distribution of gravel-bed rivers

269

distribution of the feed. As in the case of Fig. 10.9, the plot illustrates the progression from nearly static armor, to mobile armor and then to a state at which the bed is nearly unarmored as sediment feed rate qbTf increases. Figs. 10.23 and 10.24 pertain solely to the hydrograph runs. Fig. 10.23 shows plots of surface geometric mean sizes at the maximum (peak) and minimum (end) ﬂows Dsgp and Dsge, respectively, as well as geometric mean sizes of the load at the maximum (peak) and minimum (end) ﬂows Dlgp and Dlge, respectively, versus sediment feed rate qbTf. Again, the values are for the node farthest downstream and the last hydrograph of the run. Also included is the geometric mean size Dlgf of the feed sediment. Note that Dsgp and Dsge are virtually identical except at the highest feed rates, where (as is shown below) the hydrograph boundary layer becomes more diffuse. Even at the highest feed rates, however, they differ little. Note also, however, that Dlgp is always coarser than Dlge. The difference is modest for the highest feed rate qbTf ¼ 1 101 m2/s, and reaches a maximum at a feed rate near qbTf ¼ 2 104 m2/s. The difference then declines with decreasing qbTf. The decline in the difference between the bedload geometric mean sizes at the peak and end ﬂows as qbTf declines below 2 104 m2/s appears to be inherent in the formulation of Wilcock and Crowe (2003). In the case of the lowest feed rate in Fig. 10.23, the surface geometric mean size Dsgp is seen to be slightly ﬁner than that of the feed sediment. This likely reﬂects the fact that at a feed rate qbTf as low as 1 106 m2/s even 120,000 years of run time is not quite sufﬁcient to reach a (barely) mobile-bed equilibrium.

Dsgp, Dsge, Dlgp, Dlge, Dlgf (all mm)

100

geometric mean size of:

surface, peak and end flows of hydrograph Dsgp Dsge Dlgp Dlge Dlgf

feed

10

load, peak flow of hydrograph

load, end flow of hydrograph 1 1.E-06

1.E-05

1.E-04

1.E-03

1.E-02

1.E-01

qbTf (m2/s) Figure 10.23. Plots of surface and bedload geometric mean sizes Dsgp and Dlgp, respectively, at maximum (peak) ﬂow, and surface and bedload geometric mean sizes Dsge and Dlge, respectively, at minimum (end) ﬂow for the hydrograph runs of Table 10.1. The values pertain to mobile-bed equilibrium at the farthest node downstream (x ¼ L). Also shown for reference is the geometric mean size of the feed sediment Dlgf.

G. Parker, M.A. Hassan, P. Wilcock

270

percent sand in:

Fssp, Fsse, Fslp, Fsle, Fslf (all x 100)

100

load, end flow of hydrograph

feed

10

Fssp Fsse Fslp Fsle Fslf

load, peak flow of hydrograph

surface, peak and end flows of hydrograph

1 1.E-06

1.E-05

1.E-04

1.E-03

1.E-02

1.E-01

2

qbTf (m /s) Figure 10.24. Plots of fraction sand in the surface and bedload Fssp and Fslp, respectively, at maximum (peak) ﬂow, and fraction sand in the surface and bedload Fsse and Fsle at minimum (end ﬂow) for the hydrograph runs of Table 10.1. The values pertain to mobile-bed equilibrium at the farthest node downstream (x ¼ L). Also shown for reference is the fraction sand Fslf in the feed sediment.

Fig. 10.24 shows the fraction of sand in the surface and load at the maximum (peak) ﬂow Fssp and Fslp, respectively, the corresponding fractions Fsse and Fsle at the minimum (end) ﬂow, respectively, and the fraction of sand Fslf in the feed, all as functions of qbTf. Again, the values correspond to the node farthest downstream and the last hydrograph of the run. Note that the sand fractions Fssp and Fsse in the surface layer are nearly identical to each other, whereas the sand fraction in the load at peak ﬂow Fslp tends to be markedly lower than the value Fsle at the end ﬂow.

12.

A simpler model for the hydrograph boundary layer

The numerical model described above serves to identify the hydrograph boundary layer and the tradeoff associated with it, but it does not provide a particularly lucid explanation for the existence of the hydrograph boundary layer. In this section this explanation is pursued in the context of a simpler model using uniform sediment. The model is developed with the aid of singular perturbation techniques applied to boundary layer analysis (Nayfeh, 1993). Consider the conﬁguration of Fig. 10.25, which differs from that of Fig. 10.4 only in that the sediment fed in at the upstream is uniform with size D. The volume bedload transport rate for uniform sediment is denoted as qb, and the corresponding feed rate is denoted as qbf. Again, a cyclic hydrograph qw(t) is imposed, whereas the

Adjustment of the bed surface size distribution of gravel-bed rivers

271

qb initial bed profile

mobile-bed equilibrium profile (averaged over hydrograph)

Qw

η

L Figure 10.25. Schematic diagram of the conﬁguration for the numerical experiments with uniform sediment.

water surface high flow

qb

"boundary layer" water surface low flow η

qw

L

qb qw

qb η

η

t

t

Figure 10.26. Diagram illustrating the expected behavior for numerical experiments with uniform sediment. At the upstream end of the reach the volume bedload feed rate per unit width qb is held constant, but ﬂow discharge per unit width qw is allowed to vary cyclically. In a short hydrograph boundary layer downstream bed elevation Z varies cyclically as well. Downstream of this boundary layer Z becomes invariant with time, and qb now varies cyclically with the ﬂow hydrograph. The diagram pertains to mobile-bed equilibrium.

bedload feed rate qbf is held constant. Fig. 10.26 for uniform sediment is an analog of Fig. 10.10 for mixtures. It suggests that under the imposed conditions, bed elevation Z should ﬂuctuate with discharge in a short hydrograph boundary layer downstream of the feed point. Farther downstream, however, the analogous tradeoff should result in bed elevation Z that is constant in time and a bedload transport rate that ﬂuctuates over the hydrograph. Here a rescaling is used to establish the existence of this hydrograph boundary layer and quantify its characteristics. Before pursuing the details of the analysis, it is useful to describe the results in advance without the use of equations. Any river reach with a given length and

G. Parker, M.A. Hassan, P. Wilcock

272

given characteristic bed material sediment supply has a characteristic time for morphodynamic evolution in response to imposed change. Now impose a cycled ﬂow hydrograph but constant sediment feed rate onto this reach. If the time duration of the hydrograph is sufﬁciently long compared to the response time of the reach, the ﬂow hydrograph imposes cyclic aggradation and degradation throughout the reach, even at mobile-bed equilibrium. If the time duration of the hydrograph is very short compared to the response time of the reach, however, the ﬂow discharge changes so fast that the effect of any given ﬂow discharge does not have enough time to propagate very far downstream before the discharge changes. As a result bed elevation ﬂuctuations are restricted to a short hydrograph boundary layer near the feed point. Downstream of this hydrograph boundary layer bed elevation becomes invariant in time over the hydrograph, and instead the effect of the hydrograph is imprinted on the load, which varies cyclically in time. In accordance with Fig. 10.25, sediment is fed into a reach of length L at constant volume rate per unit width qbf. Flow discharge per unit width qw varies, however, according to a cyclically repeated hydrograph: t (10.23) qw ¼ q w 0 1 þ f w Th where qw0 denotes a characteristic discharge of a base state (without ﬂuctuations) and Th denotes the duration of the hydrograph. Eventually a ﬁnal equilibrium state is obtained. What are its characteristics? The Exner equation of sediment continuity for uniform sediment is written as @Z @q (10.24) ¼ b @x @t A very simple sediment transport equation is used here for illustrative purposes; ð1 lp Þ

q ¼ ab qw S

(10.25)

where ab is a constant. Such an equation can be obtained from a relation of the type of Meyer-Peter and Mu¨ller (1948) at sufﬁciently high Shields number, the assumption of a constant coefﬁcient of bed resistance Cf, where 1=2

Cf

¼

U q ¼ w un Hun

(10.26)

and the normal ﬂow assumption of (10.14), in which case 1=2

Cf (10.27) R It should be pointed out, however, that the general nature of the results obtained here are not dependent on the precise nature of the bedload transport equation. The boundary conditions on (10.24) are ab ¼ 8

qb jx¼0 ¼ qbf ;

Zjx¼L ¼ 0

(10.28a,b)

In the analysis of this section the hydrograph described by (10.23) is repeated cyclically with no intermittency, i.e. no periods of low ﬂow. That is, time is compressed to exclude periods of morphodynamically inactive low ﬂow.

Adjustment of the bed surface size distribution of gravel-bed rivers

273

The base equilibrium state associated with constant discharge qw0 and sediment feed rate qbf is given as 1 qbf ab q w x Z0 ¼ S 0 L 1 L Now the solution of (10.24) subject to (10.28) is written as S0 ¼

Z ¼ Z0 ðxÞ þ Zd ðxÞ

(10.29a) (10.29b)

(10.30)

where Zd denotes a deviatoric bed elevation around the base equilibrium state. Substituting (10.30) into (10.18) and using the result and (10.29a) to reduce (10.25), it is found that 1 @Zd (10.31) qb ¼ qbf ð1 þ f w Þ 1 S 0 @x 1 @Zd S ¼ S0 1 (10.32) S0 @x Equation (10.24) reduces with (10.29a) to ð1 lp Þ

@Zd qbf @2 Z ¼ ð1 þ f w Þ 2d @t S0 @x

The boundary conditions (10.28a,b) on (10.33) reduce to @Zd fw ¼ S0 ; Zd jx¼L ¼ 0 @x x¼0 1þ fw

(10.33)

(10.34a,b)

The problem is made dimensionless as follows: x ¼ Lxn

(10.35a)

Zd ¼ S0 LZn

(10.35b)

t ¼ ð1 lp Þ

S 0 L2 tn qbf

(10.35c)

S ¼ S0 Sn

(10.35d)

qb ¼ qbf qn

(10.35e)

Note that according to the above transformations, dimensionless distance xn varies from 0 to 1 and dimensionless slope and bedload transport rate ﬂuctuate around unity. This rescaling allows the determination of a dimensionless parameter that must be small in order for a hydrograph boundary layer to be manifested. The parameters xn and Zn are hereby termed ‘‘outer variables,’’ in the sense that they are the relevant parameters for describing channel morphodynamics outside the hydrograph boundary layer. That is, downstream distance x is normalized against total reach length L and deviatoric bed elevation Zd is normalized against the

G. Parker, M.A. Hassan, P. Wilcock

274

elevation difference between the upstream and downstream end of the reach in the absence of elevations. Note also that t is normalized against a characteristic time for morphodynamic response Tm of the entire channel reach (with length L) by aggradation or degradation, where T m ¼ ð1 lp Þ

S 0 L2 qbf

(10.35f)

so that (10.35c) can be recast in the form t ¼ T m tn

(10.35g)

That (10.35f) does indeed correspond to a characteristic morphologic response time can be seen as follows. Consider a reach of zero slope and length L. If sediment were fed into this reach at rate qbf to form wedge-shaped deposit with slope S0, vanishing elevation at the downstream end and deposit porosity lp, and no sediment outﬂow were allowed, the amount of time required to ﬁll with wedge would be equal to Tm/2. Substituting (10.35a–e) into (10.33), (10.34a,b), (10.31) and (10.32), the following dimensionless relations are obtained; @Zn @2 Z ¼ ð1 þ f w Þ 2n @tn @xn @Zn fw ¼ @xn xn ¼0 1 þ f w Zn xn ¼1 ¼ 0 Sn ¼ 1

@Zn @xn

@Z qn ¼ ð1 þ f w Þ 1 n @xn

(10.36a)

(10.36b) (10.36c) (10.37a) (10.37b)

Recalling that fw is a function of t/Th, where Th denotes the length (period) of the hydrograph, it follows from (10.35c) that t n (10.38) fw ¼ fw where ¼

qbf T h Th ¼ 2 Tm ð1 lp ÞS0 L

(10.39)

The rest of the analysis is based on the assumption that ¼

Th 1 Tm

(10.40)

This assumption implies that the duration of a single hydrograph is very short compared to the characteristic time needed to effect morphodynamic change over the

Adjustment of the bed surface size distribution of gravel-bed rivers

275

entire channel. It will be shown that a well-deﬁned hydrograph boundary layer appears under the constraint of (10.40), or more precisely when 1=2 1. The equations are further transformed from the time variable tn to a time variable t^ deﬁned so that t^ varies by a factor of unity over a single hydrograph: t^ ¼

tn t ¼ Th

(10.41)

Substituting (10.41) into (10.36a) and reducing, @Zn @2 Z ¼ ð1 þ f w Þ 2n @xn @t^

(10.42)

Now in general (10.42) combined with (10.40) implies that bed elevation is changing only slowly in time. However, when the ﬁnal (cyclic) equilibrium is reached, (10.42) combined with (10.36c) gives in the limit as e-0 the result Zn ¼ 0

(10.43)

That is, the deviatoric bed elevation (and thus deviatoric bed slope) can be set equal to zero at the ﬁnal cyclic equilibrium, and the dimensioned bed proﬁle is given by (10.29b) of the base state. At this same cyclic equilibrium the (dimensioned) sediment transport rate is given from (10.31) and (10.37b) in the limit as e-0 as qb ¼ qbf ð1 þ f w Þ

(10.44)

That is, deviatoric bed elevation and slope vanish, whereas the sediment transport rate follows the hydrograph. The above solution is precisely that hypothesized in Fig. 10.26 to prevail everywhere except in a hydrograph boundary layer near the sediment feed point; bed elevation and bed slope do not vary over the hydrograph, and the bedload transport rate tracks the hydrograph in accordance with (10.44). Now the above solution is incapable of satisfying (10.36b) at xn ¼ 0. This does not mean that the solution is wrong, but rather that it breaks down in the vicinity of xn ¼ 0. More specifically, the satisfaction of (10.36b) requires the existence of a ‘‘thin’’ boundary layer near xn ¼ 0 where the solution differs form the one given above. To this end, the transformations Zn ¼ 1=2 Z~

(10.45a)

xn ¼ 1=2 x~

(10.45b)

are introduced into (10.36a) and (10.36b), resulting in the relations @~Z @2 Z~ ¼ ð1 þ f w Þ 2 @t^ @x~ @~Z fw ¼ 1 þ fw @x~ x¼0 ~

(10.46a)

(10.46b)

The variables with the tildes in (10.45a) and (10.45b) are referred as ‘‘inner variables,’’ as they scale the problem within the hydrograph boundary layer.

276

G. Parker, M.A. Hassan, P. Wilcock

One more boundary condition on the above set is obtained by limit matching to the outer solution; ¼0 Z~ jx¼1 ~

(10.46c)

In practical terms this is replaced with Z~ jx¼ (10.46d) ~ L~ ¼ 0 where L~ might be equal to 5 or 10. (The precise value is irrelevant as long as it is sufﬁciently large compared to unity; see Nayfeh, 1993). Note that the dimensioned distance xL~ from the feed point corresponding to any value L~ is given from (10.45b) ~ so that if e1/2 is sufﬁciently small x ~ is only a small fraction of and (10.35a) as 1=2 LL, L reach length L. Note that in the above formulation f w ðt^Þ is periodic with period 1. All that remains here is for (10.46a) to be solved subject to (10.46b) and (10.46d) at cyclic equilibrium. The problem is linear; here a numerical formulation is used. In a numerical formulation, it is necessary to ‘‘spin up’’ to the mobile-bed equilibrium solution; an appropriate initial condition is Z~ jt^¼0 ¼ 0 Dimensionless load takes the following form in inner variables; @~Z qn ¼ ð1 þ f w Þ 1 @x~

(10.47)

(10.48)

Calculations were performed using a simple sinusoidal speciﬁcation for fw: f w ðt^Þ ¼ ah sinð2pt^Þ

(10.49)

where ah is a dimensionless amplitude of discharge ﬂuctuation. That is, (10.49) speciﬁes the ﬂuctuating part of the cyclically repeated hydrograph. Thus (10.46a) was solved numerically subject to the boundary conditions (10.46b) and (10.46d), the initial condition (10.47), the speciﬁcation L~ ¼ 10

(10.50) and the range 0.05oaho0.6. In performing the calculations the domain 0 x~ L~ was divided into 80 intervals and the hydrograph was discretized into 32 time steps. The calculation was continued until a ﬁnal (ﬂuctuating) equilibrium state was obtained. Fig. 10.27 shows a plot of dimensionless deviatoric bed elevation proﬁles in terms of the inner variable form Z~ versus dimensionless distance from the feed point, again in terms of the inner variable form x~ for all 32 steps of the last hydrograph of the numerical run, by which time a mobile-bed equilibrium had been achieved. Devi~ atoric bed elevation is seen to ﬂuctuate strongly in the range xo2, but for larger values of x~ the ﬂuctuations disappear. The zone where Z~ varies strongly over the hydrograph corresponds to the hydrograph boundary layer. ~ again Fig. 10.28 shows a plot of dimensionless bedload transport rate qn versus x, for all 32 steps of the ﬁnal hydrograph. Note that qn is precisely equal to 1 (bedload transport rate ¼ feed rate) at x~ ¼ 0, but downstream of x~ ¼ 2 it is seen that qn strongly tracks the hydrograph, precisely as postulated in Fig. 10.25.

Adjustment of the bed surface size distribution of gravel-bed rivers

277

0.35 0.3 0.25 0.2 0.15

~η

0.1 0.05 0 -0.05 -0.1 -0.15 -0.2

0

2

4

~ x

6

8

10

Figure 10.27. Sample calculation for uniform sediment at mobile-bed equilibrium, showing the variation of dimensionless deviatoric bed elevation Z~ (inner variable) with dimensionless distance x~ (inner variable) over the complete cycle of the last hydrograph of the run. Note that bed elevation ﬂuctuations are restricted to a hydrograph boundary layer within which x~ is less than about 2.

Dimensionless boundary layer thickness x~ d can be deﬁned as follows. Let x~ d denote the smallest value of x~ such that the following condition is satisﬁed; max j~Zðx~ d ; t^Þj wtol max j~Zð0; t^Þj

(10.51)

where wtol is a dimensionless tolerance that must take a value sufﬁciently small ~ x~ d . Here wtol is compared to unity so as to allow the neglect of ﬂuctuations in Z~ for x4 set equal to 0.01. Fig. 10.29 shows a plot of x~ d versus ah for the calculations performed here. It is seen that x~ d does not vary strongly in ah over the range, 0.05rahr0.6, ranging from about 2.5 at the lower value of ah to about 2.2 at the higher value. It should be noted that a tolerance wtol of 0.01 quite small, resulting in an estimate of boundary layer thickness that is toward the high side. It is now possible to characterize the hydrograph boundary layer thickness in dimensioned terms. Let x ¼ d denote the distance from the feed point (x ¼ 0) to a point downstream of which elevation ﬂuctuations can be neglected, i.e., the point where the equality in (10.51) is satisﬁed. This value of d denotes the thickness, or length of the hydrograph boundary layer. Between (10.35a) and (10.45b) it is found that d ¼ 1=2 x~ d L

(10.52)

G. Parker, M.A. Hassan, P. Wilcock

278 1.8 1.6 1.4 1.2

qn

1 0.8 0.6 0.4 0.2 0 0

1

2

3

4

5

6

7

8

9

10

~

X

Figure 10.28. Sample calculation for uniform sediment at mobile-bed equilibrium, showing the variation of dimensionless bedload transport rate qn with dimensionless distance x~ (inner variable) over the complete cycle of the last hydrograph of the run. Note that qn is held constant at the feed point (x~ ¼ 0), but varies cyclically with the hydrograph farther downstream. This cyclic variation in time becomes invariant in x~ for x~ greater than about 2.

Since x~ d is in the range 2.1–2.5 for the calculations performed here, it follows that the hydrograph boundary layer length is short compared to the reach length L as long as 1=2 1 Reducing (10.52) with (10.39), it is further found that " #1=2 1=2 d qbf T h Th ~ x x~ d ¼ ¼ d 2 Tm L ð1 lp ÞS 0 L

(10.53)

(10.54)

The above relation justiﬁes the verbal description given at the beginning of this section. That is, as long as hydrograph duration Th is sufﬁciently short compared to the characteristic time of morphologic response Tm, bed elevation ﬂuctuations are restricted to a short hydrograph boundary layer downstream of the feed point. Downstream of this bed elevation becomes invariant, and instead the bedload transport rate ﬂuctuates cyclically with the hydrograph. The above equation indicates that the ratio of the hydrograph boundary layer length to the reach length increases with increasing bedload transport rate qbo( ¼ feed rate). This is why at least some weak inﬂuence of the upstream conditions are felt at the downstream end of the runs with the highest bedload feed rates in Table 10.1.

Adjustment of the bed surface size distribution of gravel-bed rivers

279

2.6 2.5 2.4

~ xδ

2.3 2.2 2.1 2 0

0.1

0.2

0.3

0.4

0.5

0.6

ah Figure 10.29. Plot of dimensionless hydrograph boundary layer thickness x~ d as a function of dimensionless hydrograph amplitude for the case of uniform sediment. A very small tolerance wtol of 0.01 was used for these calculations.

Now (10.54) does not apply precisely to the runs of Table 10.1, because its derivation uses a highly simpliﬁed bedload transport relation that does not treat mixtures. This notwithstanding, the result of a more detailed analysis is likely to have a form similar to (10.54). With this in mind, (10.54) is applied as a crude approximation to estimate the ratio d/L as a function of volume bedload feed rate per unit width qbTf for the numerical runs of Table 10.1. In performing the calculation, Th is set equal to 4.5 days and lp is set equal to 0.4, as was done in the case of the numerical runs of Table 10.1. In addition, x~ d is loosely estimated as 2.5, and the value used for S0 in (10.54) is the ﬁnal equilibrium bed slope (downstream of the hydrograph boundary layer) for each run, as listed in Table 10.1 Fig. 10.30 shows the estimate of d/L versus qbTf obtained in this way for the experiments of Table 10.1. The plot suggests that the hydrograph boundary layer should be very thin indeed near the lowest feed rates, but should no longer be short compared to the reach length (and indeed, occupy on the order of 40 percent of the reach length) at the highest feed rate. This is in accord with the results presented in regard to the numerical runs of Table 10.1, according to which the hydrograph boundary layer can be expected to become thicker and more diffuse as bedload feed rate increases. The above analysis can be generalized in a straightforward way for (a) more realistic bedload transport equations for uniform sediment and (b) bedload transport equations for mixtures. Such an analysis would reveal two characteristic morphodynamics time scales for the case of mixtures, i.e. the one expressed in (10.35f) which characterizes the response time for bed aggradation or degradation, and a shorter time scale characterizing the response time for the bed surface layer to adjust to the ﬂow regime. The time scale of the hydrograph must in principle be short compared to

G. Parker, M.A. Hassan, P. Wilcock

280 0.45 0.4 0.35

δ/L

0.3 0.25 0.2 0.15 0.1 0.05 0 1.00E-06

1.00E-05

1.00E-04 1.00E-03 qbTf (m2/s)

1.00E-02

1.00E-01

Figure 10.30. Crude estimate of the variation of the ratio d/L of hydrograph boundary layer thickness to reach length versus volume sediment feed rate per unit width qbTf for the hydrograph runs of Table 10.1, based on the analysis for uniform sediment.

the smallest of these time scales in order for a distinct hydrograph boundary layer to be manifested. The two time scales reduce to a single scale for the case of uniform sediment considered in this section. The simplicity of this case allows a relatively clear explanation of the phenomenon of the hydrograph boundary layer. A more complete version of the above analysis, along with a comparison with experimental data can be found in Wong and Parker (in press).

13.

Caveats

A careful perusal of Fig. 10.19 reveals that the geometric mean size of the bedload Dlg averaged over the hydrograph at mobile-bed equilibrium is quite close to the geometric mean size of the bedload feed Dlgf, but not identical to it. The two should in fact be identical. The discrepancy is partly associated with numerical issues, particularly in the case of Run 11H, for which the feed rate qbTf is so low that even 120,000 year of run time is not quite sufﬁcient to reach mobile-bed equilibrium. There is, however, another reason for the modest discrepancy. Equation (10.9) was not precisely implemented in the calculations reported here. In a precise implementation, new substrate with a vertically varying grain size distribution would be stored as the bed aggrades, and this substrate would be mined as the bed degrades subsequently. Such an implementation, while not difﬁcult in principle, imposes large constraints on computational memory, as the vertical structure of the substrate must be stored dynamically. Here the calculations were simpliﬁed by assuming that the substrate size distribution is always equal to the initial value. This simpliﬁcation introduces a modest but discernible error in the calculations in the approach to mobile-bed equilibrium.

Adjustment of the bed surface size distribution of gravel-bed rivers

281

In several places in the description the bed is characterized as being ‘‘unarmored’’ or ‘‘nearly unarmored’’ for sufﬁciently high gravel supply. A more precise description is that the surface size distribution approaches that of the load at sufﬁciently high transport rates. This translates into the elimination of the armor layer only when the substrate has a similar size distribution, or is coarser than the load. This condition appears, however to prevail in most cases of interest (e.g., Lisle, 1995). The conclusions presented here, and more precisely the conclusion that the surface layer of a gravel-bed river tends to evolve so that its grain size distribution becomes independent of ﬂow, is based on a 1D numerical model in which all form drag and all 2D effects such as bars and local sorting have been neglected. In addition, the hydrograph is deterministic and cyclically repeated rather than stochastic. Including such factors might change the results, which thus must remain somewhat tentative. The present results apply to mobile-bed equilibrium. They may apply to ﬁeld gravel-bed streams that are (a) subject to hydrologic regime that does not deviate too much from statistical invariance and (b) are otherwise not too far from grade. They are unlikely to apply to streams that are measurably out of equilibrium.

14.

Conclusions

A 1D numerical model is presented for the morphodynamic evolution to mobile-bed equilibrium of a reach of a gravel-bed river transporting gravel and sand as bedload. A cycled hydrograph is imposed on the reach, but the upstream sediment feed rate and size distribution are held constant. The following results are obtained at mobilebed equilibrium over a broad range of conditions.

Just downstream of the feed point the bed elevation, bed slope and bed surface size distribution ﬂuctuate cyclically over the hydrograph, as the bed responds to the imbalance between transport capacity and supply. This behavior is, however, restricted to a short reach, or hydrograph boundary layer downstream of the feed point. Downstream of the hydrograph boundary layer, a tradeoff in ﬂuctuations between bed and bedload takes place. That is, the bed evolves to a bed elevation, bed slope and bed surface size distribution that become invariant to the hydrograph. The response to the hydrograph is instead expressed in terms of a bedload transport rate and a bedload size distribution that ﬂuctuate cyclically with the hydrograph. This behavior prevails over the great majority of the reach in question. The hydrograph boundary layer is shortest and sharpest when the duration of the hydrograph is very short compared to the characteristic morphodynamic response time of the reach. This is because the ﬂow changes so rapidly in time that the effect of constant bedload magnitude and size distribution imposed upstream is not able to propagate far downstream. As the duration of the hydrograph becomes longer compared to the characteristic morphodynamic response time of the reach, the hydrograph boundary layer becomes thick compared to the reach length, and more diffuse as well. Such conditions are approached as the bedload feed rate becomes sufﬁciently high.

G. Parker, M.A. Hassan, P. Wilcock

282

A hydrograph boundary layer could be identiﬁed in all cases over the range of bedload feed rates studied here. Some weak effect of the upstream boundary condition was exerted on the bed surface size distribution at the downstream end, however, in the cases of the highest bedload feed rates.

The main conclusion of this paper may have an important practical application. Let us assume, at least tentatively, that real gravel-bed rivers, with all their complications which are not included in the present numerical model, nevertheless behave similarly to that described above. It then follows that the surface size distribution present during ﬂoods (which is not easily sampled) may differ little from that prevailing at low ﬂow (which is easily sampled). If this were to be true, it would greatly simplify the application of surface-based gravel transport relations to rivers at ﬂood ﬂows.

Notations ah au B Cf D Di Dlg Dlge Dlgf Dlgp Dl50f Dsg Dsga Dsge Dsgp Ds90 Fsl Fsle Fslf Fslp Fss Fsse Fssp fh

Dimensionless amplitude of ﬂow discharge ﬂuctuation for the sinusoidal hydrograph of the analysis for uniform sediment upwinding coefﬁcient used in calculating spatial derivatives of parameters involving bedload transport channel width bed friction coefﬁcient deﬁned by (10.26) sediment size characteristic sediment size of the ith grain size range geometric mean size of bedload geometric mean size of bedload at end ﬂow of hydrograph geometric mean size of feed sediment geometric mean size of bedload at peak ﬂow of hydrograph median size of the feed sediment geometric mean size of sediment in surface layer geometric mean size of sediment in surface layer averaged over hydrograph geometric mean size of sediment in surface layer at end ﬂow of hydrograph geometric mean size of sediment in surface layer at peak ﬂow of hydrograph surface size such that 90 percent of a sample is ﬁner fraction of sand in bedload fraction of sand in bedload at end ﬂow of hydrograph fraction of sand in feed sediment fraction of sand in bedload at peak ﬂow of hydrograph fraction of sand in surface layer fraction of sand in surface layer at end ﬂow of hydrograph fraction of sand in surface layer at peak ﬂow of hydrograph functional notation for the ﬂuctuating part of the hydrograph used in the analysis for uniform sediment; see (10.23)

Adjustment of the bed surface size distribution of gravel-bed rivers fi fIi Fi g H k ks L L~

La M na nk nstep pbi qb qbf qbfi qbi qbT qbTf qn Qw qw qw 0 R S SI Sn S0 Tdur Th Tm t tn t^ U

283

volume fraction of substrate material in the ith grain size range volume fraction of material in the ith grain size range exchanged across the surface-substrate interface as the bed aggrades or degrades volume fraction of surface material in the ith grain size range gravitational acceleration ﬂow depth index for nodes in spatial discretization bed roughness height for resistance calculations reach length appropriately large value of x~ chosen to be well outside the hydrograph boundary layer (but still only a short distance downstream of the feed point compared to reach length L) thickness of the active (surface) layer number of spatial intervals in discretization coefﬁcient relating the thickness of the active (surface) layer La to surface size Ds90 coefﬁcient relating the roughness height ks of the ﬂow to surface size Ds90 number of time steps (each with length Dtf) that a single step of the hydrograph is divided into fraction of bedload material in the ith grain size range volume transport rate of bedload per unit width for uniform sediment feed value of qb volume feed rate of bedload per unit width in the ith grain size range volume transport rate of bedload per unit width in the ith grain size range volume transport rate of bedload per unit width summed over all grain sizes volume feed rate of bedload per unit width summed over all grain sizes dimensionless bedload transport rate deﬁned by (10.35e) water discharge water discharge per unit width water discharge per unit width in the absence of discharge ﬂuctuation (base equilibrium state) in the analysis for uniform sediment ¼ (rs/r)–1; submerged speciﬁc gravity of sediment bed slope initial bed slope normalized bed slope deﬁned by (10.35d) bed slope at the base equilibrium state in the analysis for uniform sediment; see (10.29a) duration of a numerical calculation hydrograph duration in the analysis for uniform sediment characteristic time for morphologic response in the analysis for uniform sediment; see (10.40) time dimensionless time deﬁned by (10.35c) dimensionless time deﬁned by (10.41) depth-averaged ﬂow velocity

284

G. Parker, M.A. Hassan, P. Wilcock

un W ni x xn x~ x~ d

shear velocity dimensionless sediment transport rate deﬁned by (10.16) distance downstream dimensionless ‘‘outer’’ streamwise distance deﬁned by (10.35a) dimensionless ‘‘inner’’ streamwise distance deﬁned by (10.45b) thickness of the hydrograph boundary layer in terms of the inner streamwise coordinate coefﬁcient in relation (10.9) for interfacial exchange fractions as bed aggrades or degrades coefﬁcient in the simple bedload relation (10.25) used in the analysis for uniform sediment coefﬁcient in Manning-Strickler resistance relation (10.11) tolerance factor used to determine the thickness of the hydrograph boundary layer in (10.51) thickness, or length downstream of x ¼ 0, of the hydrograph boundary layer temporal step length used in calculations for constant ﬂow; corresponds to actual time over a year rather than just ﬂood time temporal step length during ﬂood ﬂow for calculations using hydrographs spatial step length ratio of hydrograph duration Th to characteristic time for morphodynamic response Tm; see (10.39) bed elevation deviatoric bed elevation deﬁned by (10.30) dimensionless ‘‘outer’’ deviatoric bed elevation deﬁned by (10.35b) dimensionless ‘‘inner’’ deviatoric bed elevation deﬁned by (10.45a) bed elevation for the base equilibrium state in the analysis for uniform sediment; see (10.29b) deviatoric bed elevation deﬁned by (10.30) porosity of bed deposit water density sediment density boundary shear stress

a ab ar wtol d Dt Dtf Dx e Z Zd Zn Z~ Z0 Zd lp r rs tb

Acknowledgements This work was supported by the National Science Foundation via Agreement Number EAR-0207274. Additional support was derived from the STC program of the National Science Foundation via the National Center for Earth-surface Dynamics under Agreement Number EAR-0120914. This paper represents a contribution of the research of the National Center for Earth-surface Dynamics in the area of channel dynamics. The following graduate students in a class on morphodynamics offered by the ﬁrst author in 2004 presented as part of a ﬁnal examination preliminary versions of some of the results presented here: P. Chatanantavet, R. Hauck, W.

Adjustment of the bed surface size distribution of gravel-bed rivers

285

Kim, J.W. Lauer, N. Strong, M. Tal and M. Wong. The ﬁrst author thanks M. Wong for many interesting and helpful discussions.

References Andrews, E.D., Erman, D.C., 1986. Persistence in the size distribution of surﬁcial bed material during an extreme snowmelt ﬂood. Water Resour. Res. 22, 191–197. Hoey, T.B., Ferguson, R.I., 1994. Numerical simulation of downstream ﬁning by selective transport in gravel bed rivers: model development and illustration. Water Resour. Res. 30, 2251–2260. Hollingshead, A.B., 1971. Sediment transport measurements in a gravel river. J. Hydraul. Eng. 97 (11), 1817–1834. Hunziker, R., Jaeggi, M.N.R., 2002. Grain sorting processes. J. Hydraul. Eng. 128 (12), 1060–1068. Larson, G., 1984. In Search of the Far Side. Andrews McMeel Publishing, Kansas City, 104pp. Lisle, T.E., 1995. Particle size variations between bed load and bed material in natural gravel bed channels. Water Resour. Res. 31 (4), 1107–1118. Meyer-Peter, E., Mu¨ller, R., 1948. Formulas for Bed-Load Transport. Proceedings of 2nd Congress of International Association of Hydraulic Research, Stockholm, pp. 39–64. Milhous, R.T., 1973. Sediment transport in a Gravel-bottomed stream. Ph.D. thesis, Oreg. State University, Corvallis. Nayfeh, A.H., 1993. Introduction to Perturbation Techniques. Wiley, ISBN: 0-471-31013-1, 536pp. Parker, G., 1990. Surface-based bedload transport relation for gravel rivers. J. Hydraul. Res. 28 (4), 417–436. Parker, G., 2004. 1D Sediment Transport Morphodynamics with Applications to Rivers and Turbidity Currents, e-book downloadable at http://www.ce.umn.edu/parker/morphodynamics_e-book.htm Parker, G., Dhamotharan, S., Stefan, H., 1982. Model experiments on mobile, paved gravel bed streams. Water Resour. Res. 18 (5), 1395–1408. Parker, G., Klingeman, P.C., 1982. On why gravel bed streams are paved. Water Resour. Res. 18 (5), 1409–1423. Parker, G., Sutherland, A.J., 1990. Fluvial armor. J. Hydraul. Res. 28 (5), 529–544. Powell, D.M., Reid, I., Laronne, J.B., 2001. Evolution of bedload grain-size distribution with increasing ﬂow strength and the effect of ﬂow duration on the caliber of bedload sediment yield in ephemeral gravel-bed rivers. Water Resour. Res. 37 (5), 1463–1474. Toro-Escobar, C.M., Parker, G., Paola, C., 1996. Transfer function for the deposition of poorly sorted gravel in response to streambed aggradation. J. Hydraul. Res. 34 (1), 35–53. Wilcock, P.R., Crowe, J.C., 2003. Surface-based transport model for mixed-size sediment. J. Hydraul. Eng. 129 (2), 120–128. Wilcock, P.R., DeTemple, B.T., 2005. Persistence of armor layers in gravel-bedded streams. Geophys. Res. Lett. 32, L08402, doi:10.1029/2004GL021772, 4pp. Wilcock, P.R., Kenworthy, S.T., Crowe, J.C., 2001. Experimental study of the transport of mixed sand and gravel. Water Resour. Res. 37 (12), 3349–3358. Wong, M., Parker, G., in press. One-dimensional modeling of morphodynamic bed evolution of a gravelbed river subject to a cycled hydrograph. J. Geophys. Res. Earth Surf., preprint downloadable at: http://cee.uiuc.edu/people/parkerg/preprints.htm

Discussion by Lynne E. Frostick The authors present an interesting 1D model of gravel-bed evolution in response to simulated ﬂood hydrographs and conclude that armor layers remain in place during ﬂood ﬂows, incoming particles from upstream replacing similar particles entrained from the bed. In this model, both the amount and caliber of the material supplied from

Adjustment of the bed surface size distribution of gravel-bed rivers

285

Kim, J.W. Lauer, N. Strong, M. Tal and M. Wong. The ﬁrst author thanks M. Wong for many interesting and helpful discussions.

References Andrews, E.D., Erman, D.C., 1986. Persistence in the size distribution of surﬁcial bed material during an extreme snowmelt ﬂood. Water Resour. Res. 22, 191–197. Hoey, T.B., Ferguson, R.I., 1994. Numerical simulation of downstream ﬁning by selective transport in gravel bed rivers: model development and illustration. Water Resour. Res. 30, 2251–2260. Hollingshead, A.B., 1971. Sediment transport measurements in a gravel river. J. Hydraul. Eng. 97 (11), 1817–1834. Hunziker, R., Jaeggi, M.N.R., 2002. Grain sorting processes. J. Hydraul. Eng. 128 (12), 1060–1068. Larson, G., 1984. In Search of the Far Side. Andrews McMeel Publishing, Kansas City, 104pp. Lisle, T.E., 1995. Particle size variations between bed load and bed material in natural gravel bed channels. Water Resour. Res. 31 (4), 1107–1118. Meyer-Peter, E., Mu¨ller, R., 1948. Formulas for Bed-Load Transport. Proceedings of 2nd Congress of International Association of Hydraulic Research, Stockholm, pp. 39–64. Milhous, R.T., 1973. Sediment transport in a Gravel-bottomed stream. Ph.D. thesis, Oreg. State University, Corvallis. Nayfeh, A.H., 1993. Introduction to Perturbation Techniques. Wiley, ISBN: 0-471-31013-1, 536pp. Parker, G., 1990. Surface-based bedload transport relation for gravel rivers. J. Hydraul. Res. 28 (4), 417–436. Parker, G., 2004. 1D Sediment Transport Morphodynamics with Applications to Rivers and Turbidity Currents, e-book downloadable at http://www.ce.umn.edu/parker/morphodynamics_e-book.htm Parker, G., Dhamotharan, S., Stefan, H., 1982. Model experiments on mobile, paved gravel bed streams. Water Resour. Res. 18 (5), 1395–1408. Parker, G., Klingeman, P.C., 1982. On why gravel bed streams are paved. Water Resour. Res. 18 (5), 1409–1423. Parker, G., Sutherland, A.J., 1990. Fluvial armor. J. Hydraul. Res. 28 (5), 529–544. Powell, D.M., Reid, I., Laronne, J.B., 2001. Evolution of bedload grain-size distribution with increasing ﬂow strength and the effect of ﬂow duration on the caliber of bedload sediment yield in ephemeral gravel-bed rivers. Water Resour. Res. 37 (5), 1463–1474. Toro-Escobar, C.M., Parker, G., Paola, C., 1996. Transfer function for the deposition of poorly sorted gravel in response to streambed aggradation. J. Hydraul. Res. 34 (1), 35–53. Wilcock, P.R., Crowe, J.C., 2003. Surface-based transport model for mixed-size sediment. J. Hydraul. Eng. 129 (2), 120–128. Wilcock, P.R., DeTemple, B.T., 2005. Persistence of armor layers in gravel-bedded streams. Geophys. Res. Lett. 32, L08402, doi:10.1029/2004GL021772, 4pp. Wilcock, P.R., Kenworthy, S.T., Crowe, J.C., 2001. Experimental study of the transport of mixed sand and gravel. Water Resour. Res. 37 (12), 3349–3358. Wong, M., Parker, G., in press. One-dimensional modeling of morphodynamic bed evolution of a gravelbed river subject to a cycled hydrograph. J. Geophys. Res. Earth Surf., preprint downloadable at: http://cee.uiuc.edu/people/parkerg/preprints.htm

Discussion by Lynne E. Frostick The authors present an interesting 1D model of gravel-bed evolution in response to simulated ﬂood hydrographs and conclude that armor layers remain in place during ﬂood ﬂows, incoming particles from upstream replacing similar particles entrained from the bed. In this model, both the amount and caliber of the material supplied from

286

G. Parker, M.A. Hassan, P. Wilcock

upstream are kept constant so that the system moves from a sediment starved to an excess sediment situation as each ﬂood progresses. This is interesting, but does not reﬂect the situation in the ﬁeld where both bedload quantity and caliber vary in response to changing supply from upstream (from both banks and bed) and competence of the stream to carry material (Reid and Frostick, 1986; Ashmore, 1991; Powell et al., 2001). This could be important since we know that material of larger caliber than the bed surface is unlikely to ﬁnd an appropriate depositional niche and is likely to ‘overpass’ in a manner similar to that described by Allen (1983). A more realistic modelling scenario might be to introduce stepped changes in both quantity and character of the sediment supply, perhaps based on ﬁeld observations from an armored stream. Would the authors like to comment on this possibility and speculate on the result.

References Allen, J.R.L., 1983. Gravel overpassing on humpback bars supplied with mixed sediment: Examples from the Lower Old Red Sandstone, Southern Britain. Sedimentology 30, 285–294. Ashmore, P., 1991. Channel morphology and bed load pulses in braided gravel bed streams. Geograﬁska Annaler Sweden 73A, 37–52. Powell, D.M., Reid, I., Laronne, J.B., 2001. Evolution of bedload grain-size distribution with increasing ﬂow strength and the effect of ﬂow duration on the caliber of bedload sediment yield in ephemeral gravel-bed rivers. Water Resour. Res. 37, 1463–1474. Reid, I., Frostick, L.E., 1986. Dynamics of bedload transport in Turkey Brook, a coarse grained alluvial channel. Earth Surf. Process. Landf. 11, 143–155.

Reply by the authors Frostick makes the following important point that was not clariﬁed in the paper. The authors are delighted with the chance to address it here. In our analysis both the sediment supply rate and the grain size distribution of the supplied sediment are held constant at the upstream end. Yet in nature both could be expected to vary. As Frostick notes, ‘‘A more realistic modeling scenario might be to introduce stepped changes in both quantity and character of the sediment supply, perhaps based on ﬁeld observations from an armored stream.’’ In our model we did not vary the supply rate and grain size distribution of the sediment supplied from upstream because we could think of no objective basis for doing so. Fortunately, however, the model makes its own decision in this regard. The model predicts the following in response to a cycled hydrograph. Within a short boundary layer reach downstream of the feed point the bed cyclically aggrades and degrades, and the bed surface cyclically coarsens and ﬁnes. Downstream of this boundary layer the bed elevation and surface size distribution become invariant, and instead the magnitude of the bedload transport rate and the bedload grain size distribution cycle with the hydrograph. Once beyond the boundary layer reach, the nature of this cycling is invariant with distance downstream. Now suppose that the precise pattern of cyclic variation of bedload transport rate and grain size distribution observed downstream of the boundary layer is used to

286

G. Parker, M.A. Hassan, P. Wilcock

upstream are kept constant so that the system moves from a sediment starved to an excess sediment situation as each ﬂood progresses. This is interesting, but does not reﬂect the situation in the ﬁeld where both bedload quantity and caliber vary in response to changing supply from upstream (from both banks and bed) and competence of the stream to carry material (Reid and Frostick, 1986; Ashmore, 1991; Powell et al., 2001). This could be important since we know that material of larger caliber than the bed surface is unlikely to ﬁnd an appropriate depositional niche and is likely to ‘overpass’ in a manner similar to that described by Allen (1983). A more realistic modelling scenario might be to introduce stepped changes in both quantity and character of the sediment supply, perhaps based on ﬁeld observations from an armored stream. Would the authors like to comment on this possibility and speculate on the result.

References Allen, J.R.L., 1983. Gravel overpassing on humpback bars supplied with mixed sediment: Examples from the Lower Old Red Sandstone, Southern Britain. Sedimentology 30, 285–294. Ashmore, P., 1991. Channel morphology and bed load pulses in braided gravel bed streams. Geograﬁska Annaler Sweden 73A, 37–52. Powell, D.M., Reid, I., Laronne, J.B., 2001. Evolution of bedload grain-size distribution with increasing ﬂow strength and the effect of ﬂow duration on the caliber of bedload sediment yield in ephemeral gravel-bed rivers. Water Resour. Res. 37, 1463–1474. Reid, I., Frostick, L.E., 1986. Dynamics of bedload transport in Turkey Brook, a coarse grained alluvial channel. Earth Surf. Process. Landf. 11, 143–155.

Reply by the authors Frostick makes the following important point that was not clariﬁed in the paper. The authors are delighted with the chance to address it here. In our analysis both the sediment supply rate and the grain size distribution of the supplied sediment are held constant at the upstream end. Yet in nature both could be expected to vary. As Frostick notes, ‘‘A more realistic modeling scenario might be to introduce stepped changes in both quantity and character of the sediment supply, perhaps based on ﬁeld observations from an armored stream.’’ In our model we did not vary the supply rate and grain size distribution of the sediment supplied from upstream because we could think of no objective basis for doing so. Fortunately, however, the model makes its own decision in this regard. The model predicts the following in response to a cycled hydrograph. Within a short boundary layer reach downstream of the feed point the bed cyclically aggrades and degrades, and the bed surface cyclically coarsens and ﬁnes. Downstream of this boundary layer the bed elevation and surface size distribution become invariant, and instead the magnitude of the bedload transport rate and the bedload grain size distribution cycle with the hydrograph. Once beyond the boundary layer reach, the nature of this cycling is invariant with distance downstream. Now suppose that the precise pattern of cyclic variation of bedload transport rate and grain size distribution observed downstream of the boundary layer is used to

Adjustment of the bed surface size distribution of gravel-bed rivers

287

specify the feed condition. If this is done, the boundary layer itself disappears. That is, at mobile-bed equilibrium the bed elevation and surface grain size distribution remain constant everywhere, and the bedload transport rate and grain size distribution cycle in the same way with the hydrograph everywhere. Outside of the boundary layer, the equilibrium attained in this way is identical to that attained using the simpler condition of constant rate of introduction and size distribution of the sediment feed. That is, in some sense the model ‘‘tells’’ the user how to vary the rate of introduction and size distribution of the sediment feed so as to obtain an equilibrium bed proﬁle in the presence of a hydrograph. The way the model works is entirely consistent with the back-calculation performed by Wilcock and DeTemple (2005). Frostick’s comment addresses one other issue, i.e., overpassing by coarse grains. The present bedload transport formulation may not be sufﬁcient to describe this interesting process.

Discussion by Matthieu de Linares The authors’ conclusion that the surface GSD is the same at low ﬂows than at high ﬂows may not be valid in the case of a river featuring rifﬂes/pools. Given that: (i) rifﬂes are generally coarser than pools, (ii) during high ﬂows bed shear stress is generally lower on the rifﬂes than on the pools (in contrary to low ﬂows), it should rather be expected that the surface GSD on the rifﬂes is ﬁner at high ﬂow and gets back to its usual coarse state during the tail of the hydrograph (reverse situation for pools). This should appear in results of the simulations, provided that (i) and (ii) are respected. That supposes a ﬁne enough representation of the topography of the rifﬂe/ pool system and a realistic initialization of the GSD.

Reply by the authors The writer indicates, ‘‘The authors’ conclusion that the surface GSD is the same at low ﬂows than at high ﬂows may not be valid in the case of a river featuring rifﬂes/ pools.’’ We suspect that the author is right. Our analysis should apply most accurately to a relatively straight gravel-bed stream with a minimum of form drag. While step-pool streams are often relatively straight, form drag is usually significant in them. In step-pool streams this form drag is often associated with the deposition of the coarsest material on the waning limb of a hydrograph. An extension of the model to this case would likely yield a difference in the surface grain size distribution as a function of ﬂow. Discussion by Ramon Batalla, Celso Garcia, and Damia´ Vericat Parker et al. suggest that during cycled hydrographs the bed-elevation and surface size distribution become invariant under constant bedload feed rate and bedload size

Adjustment of the bed surface size distribution of gravel-bed rivers

287

specify the feed condition. If this is done, the boundary layer itself disappears. That is, at mobile-bed equilibrium the bed elevation and surface grain size distribution remain constant everywhere, and the bedload transport rate and grain size distribution cycle in the same way with the hydrograph everywhere. Outside of the boundary layer, the equilibrium attained in this way is identical to that attained using the simpler condition of constant rate of introduction and size distribution of the sediment feed. That is, in some sense the model ‘‘tells’’ the user how to vary the rate of introduction and size distribution of the sediment feed so as to obtain an equilibrium bed proﬁle in the presence of a hydrograph. The way the model works is entirely consistent with the back-calculation performed by Wilcock and DeTemple (2005). Frostick’s comment addresses one other issue, i.e., overpassing by coarse grains. The present bedload transport formulation may not be sufﬁcient to describe this interesting process.

Discussion by Matthieu de Linares The authors’ conclusion that the surface GSD is the same at low ﬂows than at high ﬂows may not be valid in the case of a river featuring rifﬂes/pools. Given that: (i) rifﬂes are generally coarser than pools, (ii) during high ﬂows bed shear stress is generally lower on the rifﬂes than on the pools (in contrary to low ﬂows), it should rather be expected that the surface GSD on the rifﬂes is ﬁner at high ﬂow and gets back to its usual coarse state during the tail of the hydrograph (reverse situation for pools). This should appear in results of the simulations, provided that (i) and (ii) are respected. That supposes a ﬁne enough representation of the topography of the rifﬂe/ pool system and a realistic initialization of the GSD.

Reply by the authors The writer indicates, ‘‘The authors’ conclusion that the surface GSD is the same at low ﬂows than at high ﬂows may not be valid in the case of a river featuring rifﬂes/ pools.’’ We suspect that the author is right. Our analysis should apply most accurately to a relatively straight gravel-bed stream with a minimum of form drag. While step-pool streams are often relatively straight, form drag is usually significant in them. In step-pool streams this form drag is often associated with the deposition of the coarsest material on the waning limb of a hydrograph. An extension of the model to this case would likely yield a difference in the surface grain size distribution as a function of ﬂow. Discussion by Ramon Batalla, Celso Garcia, and Damia´ Vericat Parker et al. suggest that during cycled hydrographs the bed-elevation and surface size distribution become invariant under constant bedload feed rate and bedload size

Adjustment of the bed surface size distribution of gravel-bed rivers

287

specify the feed condition. If this is done, the boundary layer itself disappears. That is, at mobile-bed equilibrium the bed elevation and surface grain size distribution remain constant everywhere, and the bedload transport rate and grain size distribution cycle in the same way with the hydrograph everywhere. Outside of the boundary layer, the equilibrium attained in this way is identical to that attained using the simpler condition of constant rate of introduction and size distribution of the sediment feed. That is, in some sense the model ‘‘tells’’ the user how to vary the rate of introduction and size distribution of the sediment feed so as to obtain an equilibrium bed proﬁle in the presence of a hydrograph. The way the model works is entirely consistent with the back-calculation performed by Wilcock and DeTemple (2005). Frostick’s comment addresses one other issue, i.e., overpassing by coarse grains. The present bedload transport formulation may not be sufﬁcient to describe this interesting process.

Discussion by Matthieu de Linares The authors’ conclusion that the surface GSD is the same at low ﬂows than at high ﬂows may not be valid in the case of a river featuring rifﬂes/pools. Given that: (i) rifﬂes are generally coarser than pools, (ii) during high ﬂows bed shear stress is generally lower on the rifﬂes than on the pools (in contrary to low ﬂows), it should rather be expected that the surface GSD on the rifﬂes is ﬁner at high ﬂow and gets back to its usual coarse state during the tail of the hydrograph (reverse situation for pools). This should appear in results of the simulations, provided that (i) and (ii) are respected. That supposes a ﬁne enough representation of the topography of the rifﬂe/ pool system and a realistic initialization of the GSD.

Reply by the authors The writer indicates, ‘‘The authors’ conclusion that the surface GSD is the same at low ﬂows than at high ﬂows may not be valid in the case of a river featuring rifﬂes/ pools.’’ We suspect that the author is right. Our analysis should apply most accurately to a relatively straight gravel-bed stream with a minimum of form drag. While step-pool streams are often relatively straight, form drag is usually significant in them. In step-pool streams this form drag is often associated with the deposition of the coarsest material on the waning limb of a hydrograph. An extension of the model to this case would likely yield a difference in the surface grain size distribution as a function of ﬂow. Discussion by Ramon Batalla, Celso Garcia, and Damia´ Vericat Parker et al. suggest that during cycled hydrographs the bed-elevation and surface size distribution become invariant under constant bedload feed rate and bedload size

Adjustment of the bed surface size distribution of gravel-bed rivers

287

specify the feed condition. If this is done, the boundary layer itself disappears. That is, at mobile-bed equilibrium the bed elevation and surface grain size distribution remain constant everywhere, and the bedload transport rate and grain size distribution cycle in the same way with the hydrograph everywhere. Outside of the boundary layer, the equilibrium attained in this way is identical to that attained using the simpler condition of constant rate of introduction and size distribution of the sediment feed. That is, in some sense the model ‘‘tells’’ the user how to vary the rate of introduction and size distribution of the sediment feed so as to obtain an equilibrium bed proﬁle in the presence of a hydrograph. The way the model works is entirely consistent with the back-calculation performed by Wilcock and DeTemple (2005). Frostick’s comment addresses one other issue, i.e., overpassing by coarse grains. The present bedload transport formulation may not be sufﬁcient to describe this interesting process.

Discussion by Matthieu de Linares The authors’ conclusion that the surface GSD is the same at low ﬂows than at high ﬂows may not be valid in the case of a river featuring rifﬂes/pools. Given that: (i) rifﬂes are generally coarser than pools, (ii) during high ﬂows bed shear stress is generally lower on the rifﬂes than on the pools (in contrary to low ﬂows), it should rather be expected that the surface GSD on the rifﬂes is ﬁner at high ﬂow and gets back to its usual coarse state during the tail of the hydrograph (reverse situation for pools). This should appear in results of the simulations, provided that (i) and (ii) are respected. That supposes a ﬁne enough representation of the topography of the rifﬂe/ pool system and a realistic initialization of the GSD.

Reply by the authors The writer indicates, ‘‘The authors’ conclusion that the surface GSD is the same at low ﬂows than at high ﬂows may not be valid in the case of a river featuring rifﬂes/ pools.’’ We suspect that the author is right. Our analysis should apply most accurately to a relatively straight gravel-bed stream with a minimum of form drag. While step-pool streams are often relatively straight, form drag is usually significant in them. In step-pool streams this form drag is often associated with the deposition of the coarsest material on the waning limb of a hydrograph. An extension of the model to this case would likely yield a difference in the surface grain size distribution as a function of ﬂow. Discussion by Ramon Batalla, Celso Garcia, and Damia´ Vericat Parker et al. suggest that during cycled hydrographs the bed-elevation and surface size distribution become invariant under constant bedload feed rate and bedload size

G. Parker, M.A. Hassan, P. Wilcock

288

distribution. Our observations in the large regulated lower Ebro River, NE Spain, indicate that under supply-limited conditions both surface grain-size distribution and bed-elevation vary much with ﬂow (Vericat et al., 2005). Regulation took place 40 years ago but the river bed is still active. Measurements have been obtained throughout a series of ﬂoods (i.e. repeated hydrographs) of different magnitude, ranging from Q1.5 (t ¼ 39 N/m2) to Q8 (t ¼ 63 N/m2) occurred during 2002–2004. Large ﬂoods with no supply of sediment from tributaries below dams resulted in a generalized river bed incision and in a ﬁner surface layer (i.e. the armor ratio decreased considerably). Field data evidences that the surface did not remain armored during large ﬂoods, therefore low ﬂow armor layer did not persist. Bedload grain size increased with transport rates during armor breakup (December 2002 ﬂood). In contrast, they did not show any relation during subsequent ﬂoods (February and March 2003), this fact being attributed to the increment of the amount of transportable subsurface material, after the armor was locally or totally disrupted (Fig. 10.31). At the light of these observations, riverbed grain-size distribution seeing at low ﬂows may differ considerably to that at high ﬂows, making uncertain its extrapolation for modelling purposes and for bedload transport estimations by means of formula. Small ﬂoods caused no changes in bed-elevation and the recoarsening of the bed surface hence the armor reestablished. Surface deposits coarsened, increasing bed stability, minimizing incision, and decreasing bedload transport rates. Results demonstrate the strong control that ﬂood magnitude exerts on bed surface dynamics and that in the absence of inputs of sediment from upstream the surface layer absorbs an important part of the river bed variation, together with bedload.

D50-ib (mm)

50

December 2002 February 2003 March 2003

5

0.05

5.00 0.50

500.00 50.00

ib (g/ms) Figure 10.31. Bedload transport rates (ib) and median bedload grain size (D50ib) during ﬂoods in the Ebro River in 2002–2003.

Adjustment of the bed surface size distribution of gravel-bed rivers

289

Reference Vericat, D., Batalla, R.J., Garcia, C., 2005. Breakup and reestablishment of the armour layer in a large gravel-bed river below dams: the lower Ebro. Geomorphology 76, 122–136.

Reply by the authors The model presented in the paper assumes that the sediment supply is constant, either strictly in time or averaged over the hydrograph. Different results can be expected if the sediment supply is limited in a time-dependent way, or if supply varies over time scales that differ from that of a typical ﬂood hydrograph. Although not treated in the paper, the model is easily modiﬁed to treat these cases, as long as a protocol for the time variation of the magnitude and grain size distribution of the sediment delivery is known. The result of a surface grain size distribution that is invariant in time is not likely to be predicted by the model in such cases.

Adjustment of the bed surface size distribution of gravel-bed rivers

289

Reference Vericat, D., Batalla, R.J., Garcia, C., 2005. Breakup and reestablishment of the armour layer in a large gravel-bed river below dams: the lower Ebro. Geomorphology 76, 122–136.

Reply by the authors The model presented in the paper assumes that the sediment supply is constant, either strictly in time or averaged over the hydrograph. Different results can be expected if the sediment supply is limited in a time-dependent way, or if supply varies over time scales that differ from that of a typical ﬂood hydrograph. Although not treated in the paper, the model is easily modiﬁed to treat these cases, as long as a protocol for the time variation of the magnitude and grain size distribution of the sediment delivery is known. The result of a surface grain size distribution that is invariant in time is not likely to be predicted by the model in such cases.

Gravel-Bed Rivers VI: From Process Understanding to River Restoration H. Habersack, H. Pie´gay, M. Rinaldi, Editors r 2008 Elsevier B.V. All rights reserved.

291

11 Bed load transport and streambed structure in gravel streams Panos Diplas and Hafez Shaheen

Abstract Phenomena related to gravel-bed rivers continue to receive considerable attention. This effort has been facilitated by the collection of extensive ﬁeld data under relatively wide range of ﬂow, bed load transport and channel slope conditions over the last decade. In some cases, these data have provided the opportunity to go beyond bed load transport and boundary shear stress correlations and have enabled researchers to examine the behavior of gravel streams within the context of watershed processes and characteristics. Several of the gravel-bed stream studies published during the past 5 years have been analyzed and some of the information has been synthesized in an effort to attain an improved understanding about the interplay between sediment supply, armor layer development and bed load transport. Although the development of the ‘‘general equation’’ of bed load transport remains elusive, better comprehension about the complexity of the problem, given the remarkable variability of prevailing ﬁeld conditions, has been obtained and improved understanding about overall trends and interaction of the various processes has been attained. The present state of the subject provides a sound basis for signiﬁcant future developments.

1.

Introduction

During the last three decades issues related to sediment transport and the structure of channel beds in gravel streams have received considerable attention. Terms like equal mobility and size-selective transport, pavement and armor layers, fractional transport and single-diameter-based bed load transport, surface- and subsurface-based analyses have become commonplace in the gravel-bed rivers literature. The interest generated in this subject has resulted in improved methods for collecting bed load data and sampling bed material, which in turn provide much needed reliable data from ﬁeld and laboratory settings. These developments have been instrumental in the evolution of ideas in this area, which are heavily based on the interpretation of the trends exhibited by available data. As one might expect, the limited range of conditions E-mail address: [email protected] ISSN: 0928-2025

DOI: 10.1016/S0928-2025(07)11128-7

P. Diplas, H. Shaheen

292

represented by each dataset reveals part of the puzzle. However, by now, there are a sufﬁcient number of datasets to allow for the consideration of the phenomena for a relatively wide range of ﬂow, bed load transport and other important parameters. This enables us to approach the subject from a more global perspective by putting together some of the available pieces and speculate about some of the missing ones. At ﬁrst, the present study provides a brief account of some key contributions to the sediment transport ﬁeld that have had an impact in our understanding of gravel-bed stream phenomena made during the past two centuries and even earlier than that. Next, a more detailed review of some of the recent developments in bed load transport and composition of the streambed is reported. An effort to synthesize existing information on gravel streams follows. Some trends of bed load transport rates based on available ﬁeld and laboratory data covering a wide range of Shields stresses are examined and the behavior of the armor layer under changing ﬂow and sediment supply conditions is discussed.

2.

Brief historical account

Some of the ﬁrst insightful descriptions regarding gravel-bed river processes were made about 500 years ago by Leonardo, a keen observer of natural phenomena and possibly the initiator of the use of experiments as a tool in scientiﬁc inquiry. In his famous Codex Leicester he writes: ‘‘A river that ﬂows from the mountains deposits a great quantity of large stones in its bed y as it proceeds on its course it carries down with it lesser stones with the angles more worn away, and so the large stones make smaller ones; and farther on it deposits ﬁrst coarse and then ﬁne gravel, and after this follows sand at ﬁrst coarse and then more ﬁne y’’ (Richter, 1998). Leonardo attributes downstream ﬁning to abrasion and selective transport, not much different than the explanations provided by contemporary researchers. Another observation stated in the same codex, is with regard to sediment eroded in the mountains, supplied to the river, and subsequently transported to the sea, an indication about the linkage between rivers and their drainage basins. The more systematic study of rivers started during the second half of the 19th century. Until the 1970s the emphasis, for good reason, was on sandy streams. Gravel-bed streams were not perceived as behaving sufﬁciently differently from their sand-bed counterparts to deserve special attention. This perception was reﬂected in the ASCE Manual 54 (Vanoni, 1975). Nevertheless, this period was punctuated by a good number of important studies dealing with various aspects of gravel-bed streams or which indirectly beneﬁted the study of such streams. An incomplete list of some of these contributions includes: the development of a bed load transport model based on the use of tractive force (DuBoys, 1879), description of downstream ﬁning based on observations from the Alpine Rhine River (Sternberg, 1875), Gilbert’s (1914) extensive dataset based on ﬂume experiments, identiﬁcation of imbricated structures (Johnston, 1922; Lane and Carlson, 1953), use of the dimensionless boundary shear stress as a criterion for characterizing the threshold of particle motion condition (Shields, 1936), development of a bed load transport expression based on ﬂume experiments with coarse and poorly sorted sediments (Meyer-Peter and Muller, 1948),

Bed load transport and streambed structure in gravel streams

293

recognition of the ﬂuctuating nature of turbulent forces near a rough boundary and its importance in particle entrainment (Einstein and El-Samni, 1949), development of a fractional bed load transport equation that accounts for hiding effects and is not based on the notion of excess boundary shear stress (Einstein, 1950). Development of a static armor layer under starved sediment input conditions in laboratory experiments carried out by Harrison (1950), the grid-by-number method proposed by Wolman (1953) for sampling sediment deposits, especially suitable for sampling armor layers, detailed observations from Klaralven River regarding behavior of poorly sorted and bimodal sediments, among other things (Sundborg, 1956), documentation of the absence of certain grain sizes, in the pea and pebble range, from gravel-bed sediment deposits (Yatsu, 1957), reﬁnement of the threshold of motion criterion, speciﬁcally obtained for poorly sorted sediments (Pantelopulos, 1955, 1957; Neil, 1968) (middle-size fractions move at dimensionless shear stress values comparable to those required from uniform sediments), a new expression for describing hiding effects in a sediment mixture (Egiazaroff, 1965), introduction of the stream power concept in sediment transport (Bagnold, 1966), development of the Helley–Smith bed load sampler (Helley and Smith, 1971), equivalency criteria among the various methods employed for sampling bed material proposed by Kellerhals and Bray (1971). New bed load transport formulae proposed by Ashida and Michiue (1972) and Ackers and White (1973), and a high quality bed load transport dataset collected by Milhous (1973) from Oak Creek, a gravel-bed stream in Oregon, USA, using a vortex bed load sampler. During the 1980s the number of studies focusing exclusively on gravel-bedded streams grew steadily. This is partly due to the interest generated by the publications resulting from the Gravel-Bed Rivers Symposia and several key papers, including those by Parker and Klingeman (1982) and Parker et al. (1982). Extensive ﬂume and ﬁeld studies were undertaken to augment the available datasets. The interest continues unabated to the present time, to the extent that currently it has possibly surpassed the efforts devoted to the study of sandy streams. A recent search of the Web of Science database identiﬁed several hundred journal publications dealing with gravel-bed stream issues. The subject area has matured. The new ASCE Sedimentation Manual, No. 110, due to be published in 2006, will include several chapters on gravel bottomed streams, reﬂecting the explosion of knowledge acquired in this area over the last 30 years.

3.

Recent developments

Since the Gravel-Bed Rivers volumes provide extensive accounts of the developments in this area, the present brief review will focus on the highlights of the many works that were published during the last 5 years or so. Some of the topics addressed over this time period deal with bed load transport issues, headwater (high slope) streams, threshold, behavior of ephemeral streams, stream dependence on drainage basin characteristics, partial transport, and evolution of armor layer during ﬂoods. Most of these issues are intertwined. Several new bed load formulae were proposed, while some older and well-known equations were tested against new and existing ﬁeld and laboratory data. Almedeij and

294

P. Diplas, H. Shaheen

Diplas (2003) used the Oak Creek (Milhous, 1973) and Nahal Yatir (Reid et al., 1995) data, representing the two extremes, very low and very high, respectively, of bed load transport behavior to develop a two parameter equation, one based on surface and the other on subsurface material characteristics. The mode of each (surface and subsurface) population was chosen as a suitable parameter for representing the corresponding sediment characteristics. This is intended for bed load transport calculations of unimodal sediments over a wide range of boundary shear stresses. Field data from three other streams were used to examine the validity of the proposed equation. Barry et al. (2004, 2005) proposed a simple power relation for bed load transport based on total water discharge. They hypothesized that the exponent of this relation depends on the degree of channel bed armoring, while the coefﬁcient depends on basin characteristics and the amount of sediment supplied to the stream. They used a large number of ﬁeld data, collected from 24 gravel-bed rivers in Idaho, to parameterize the exponent and coefﬁcient in terms of channel and watershed characteristics, respectively. They suggested that the expression for the former parameter should be generally valid from stream to stream while the latter will be region speciﬁc, dependent upon local basin land use and physiography. Subsequently, they compared their formula, together with four other well-known bed load equations, with data obtained from 17 stream sites located in Colorado, Oregon and Wyoming. The results are typical of similar comparisons; the predicted values are within two to three orders of magnitude of the observed values. Though the results regarding the predictive ability of the new and existing equations are not very encouraging, the approach followed by the authors, to consider river processes within the context of the characteristics of their basins, appears to be in the right direction (e.g., Ryan et al., 2005). The role of sediment supply to stream behavior, in terms of degree of armoring and bed load transport, has been emphasized by many researchers (e.g., Kellerhals, 1967; Reid et al., 1985; Kuhnle and Southard, 1988; Dietrich et al., 1989; Bufﬁngton and Montgomery, 1999; Lisle et al., 2000; Emmett and Wolman, 2001; Gomi and Sidle, 2003; Whiting and King, 2003). This is demonstrated by considering the modeling results represented by sediment feed and recirculating ﬂume experiments. It is well known that these two ﬂume types provide substantially different results (e.g., Parker and Wilcock, 1993; Wilcock, 2001). The latter operates as a closed system and the results depend on the initial conditions, while the former operates as an open system and the outcome of the experiments depends on the boundary conditions (sediment feed rate) as well. Ryan et al. (2005) analyzed bed load measurements collected by the US Forest Service from steep, coarse-grained streams in Colorado and Wyoming in an effort to identify patterns exhibited by transported sediments. For their work, they selected 19 sites representing a variety of channel types, including step-pool, plane-bed and pool rifﬂe channels. Their results corroborated the two-phase bed load transport model that has been advocated in the past for gravel-bed streams (e.g., Jackson and Beschta, 1982). Phase I, dominated by low rates of sand transport, is followed by a rapid increase in both sediment size and transport rate during phase II, typically triggered by the breakup of the armor layer at near-bankfull conditions. Furthermore, the authors found that piecewise linear functions, one for each phase, provided good representation of the bed load transport data.

Bed load transport and streambed structure in gravel streams

295

Whiting and King (2003) used an extensive set of bed load transport and other channel related data collected by the US Geological Survey and the US Forest Service (King et al., 2004) at sites along 12 streams located within the Snake River Basin in Idaho over long periods to demonstrate the inﬂuence of sediment supply on the make up of the armor layer. The drainage area of the sites varied 50-fold and the channel slope ranged from 0.0005 to 0.0268. The median size of the subsurface material, D50s, ranged from 14 to 26 mm, while the corresponding values for the armor layer, D50, exhibited much wider variation, from 31 to 173 mm. The resulting degree of coarseness (D50/D50s) ranged from 1.9 to 7.2, reﬂecting substantially different rates of sediment supply from the corresponding drainage areas. Typically, the median size of the armor layer varies inversely with sediment supply, while bed load transport increases substantially with sediment supply (e.g., Kuhnle and Southard, 1988; Whiting and King, 2003). The correlation between (D50/D50s) and D* (the ratio of D50s to the median size of transport- and frequency-weighted bed load) (Lisle, 1995) based on the Whiting and King data is shown in Fig. 11.1. The degree of armoring increases as the grain size distribution of the bed load material compared with that of the subsurface decreases (r2 ¼ 0.56). It is therefore evident that decoupling the stream from its coarse sediment supply, especially for its mountainous sections, imposes a limitation on the validity of a bed load transport formula outside its database. The effects of sediment supply on stream behavior probably account for the observation reported in the literature that bed load transport formulae based on laboratory experiments (with sediment fed at the upstream end or recirculated), which do not experience sediment supply problems, often overestimate bed load transport rates collected from the ﬁeld. However, consideration of the sources supplying sediment to the stream remains a difﬁcult task. Coarse sediment supplying events, such as landslides and bank collapses, tend to be episodic and difﬁcult to model. Furthermore, they introduce considerable variability/ﬂuctuations on bed load transport for identical ﬂow conditions as they affect the availability of sediment. A surface-based transport equation, suitable for bimodal sediments, was proposed by Wilcock and Crowe (2003). The development of the sediment transport model was based on 48 ﬂume runs, performed using 5 different experimental sediments. The ﬂume operated in the sediment recirculating mode. Several aspects of the Parker 8

D50 /D50s

6 4 2 0 0

5

10

15

D* Figure 11.1. Variation of the degree of armoring with change in degree of bed load coarseness.

296

P. Diplas, H. Shaheen

(1990) surface-based bed load formula have been adopted in the new formulation. Further testing with additional ﬁeld and laboratory data will be necessary to examine the validity of this formulation. The same experiments were used to examine the role of the sand content on the mobility of a sediment mixture and the evolution of the armor layer with ﬂow strength (Wilcock et al., 2001). A substantial increase of gravel transport rate, more than an order of magnitude, with sand content was observed. However, for a given sediment mixture, no signiﬁcant change in the grain size of the surface layer with changing ﬂow strength was detected. This led them to question the traditionally held belief that the armor layer weakens with increasing ﬂow strength and transport rate. Several researchers have pointed out that high gradient streams (S40.05, S is the channel bed slope), such as many headwater streams, have certain features not commonly encountered in gravel streams with milder slopes. They include a wider range of particle sizes, sometimes extended all the way to boulders, stepped-bed morphology and shallow ﬂows compared with the coarsest sizes present on the channel bed, resulting in substantial form drag and distortion of the logarithmic velocity proﬁle. Rickenmann (2001), among many others (e.g., Gomi and Sidle, 2003; Mueller et al., 2005), pointed out that for many of these streams, ﬂow depth measurements, which can be used to determine boundary shear stress, are not usually available because they involve signiﬁcant errors. To overcome this shortcoming, he proposed a bed load transport formula based on water discharge. The empirical formula, derived from ﬂume experiments, is a linear function of the unit discharge in excess of a critical value, a result reminiscent of the formula proposed by Schoklitsch (1962). Field data from 19 mountain streams used in this study support, on average, such trends. However, the variability from stream to stream is substantial. Mueller et al. (2005) examined coupled measurements of ﬂow and bed load transport characteristics from 45 gravel streams exhibiting a wide range of channel slopes, to determine the variation of the reference bed shear stress, a stress associated with a very small but measurable bed load transport rate (e.g., Parker et al., 1982). They concluded that the dimensionless reference shear stress increases systematically with channel gradient, ranging from 0.025 at low slopes to values greater than 0.10 for slopes steeper than 0.02. Furthermore, they found that for nearly all the cases, the dimensionless bed shear stress at bankfull conditions modestly exceeded the corresponding reference value. This is consistent with conclusions reached by Parker et al. (1982) for the case of perennial gravel streams with milder slopes. The high stresses typically encountered in streams with steep slopes are moderated by excessive resistance, while smaller particles experience greater hiding effects due to the very wide range of particle sizes available on the channel bed. The need for using an appropriate skin friction value, that is the total boundary shear stress adjusted for form drag contributed by large and exposed particles, in bed load transport calculations has been emphasized by several researchers (e.g., Andrews, 2000). Behavior of gravel streams at the two ends of the spectrum, near threshold conditions and at very high shear stresses, has received considerable attention lately. Several ﬁeld studies have veriﬁed the persistence of the partial transport phenomenon in natural streams (e.g., Lisle et al., 2000; Church and Hassan, 2002; Haschenburger and Wilcock, 2003), which was initially documented in a detailed way in the

Bed load transport and streambed structure in gravel streams

297

laboratory by Wilcock and McArdell (1993). It was also reiterated that an armor layer of given grain size composition has an additional degree of freedom in enhancing the stability of its particles, when necessary. This can be accomplished by rearranging the orientation of the particles (e.g., Profﬁtt, 1980), and developing clusters and other surface structures that maximize particle resistance (e.g., Bufﬁngton and Montgomery, 1997, 1999; Church and Hassan, 2002). As a result, critical shear stress for incipient motion can vary by as much as an order of magnitude (Church and Hassan, 2002), an important consideration when bed load transport calculations are based on excess shear stress expressions. An area of fruitful research during the last decade or so has been based on the analysis and interpretation of data collected under conditions of very high bed load transport rates, typically encountered in some ephemeral streams and canyon rivers (e.g., Laronne and Reid, 1993; Reid and Laronne, 1995; Reid et al., 1998; Martin-Vide et al., 1999; Powell et al., 2001; Cohen and Laronne, 2005). These streams exhibit neutral stratiﬁcation of the bed material in terms of grain size (e.g., Powell et al., 2001; Cohen and Laronne, 2005). In the case of Nahal Yatir, Israel, though, where extreme ﬂoods generated Shields stresses as high as 13 times the critical value, a layer of ﬁner particle (with a median size of 6 mm) overlying coarser material (with a median size of 10 mm) was observed (Laronne and Reid, 1993). In all cases, the bed load material became progressively coarser with increasing shear stress, and reached equal mobility conditions at high shear stress values (44.5 the critical value; Powell et al., 2001). Though massive amounts are in transport in these streams, the bed load behavior seems to be more straightforward with fewer ﬂuctuations (Kuhnle and Southard, 1988). However, even under such conditions, there is disagreement about the validity of a suitable bed load transport formula. For example, Reid et al. (1996) determined that the Meyer-Peter and Muller (M-P & M) equation matched very well the bed load transport trends exhibited by Nahal Yatir, while Martin-Vide et al. (1999) estimated that the bed load transport rates measured from Riera de les Arenes, a steep ephemeral stream near Barcelona, Spain, were at least four times as high as those calculated from M-P & M.

4.

A bed load model that accounts for fractional sediment transport

The brief review presented in the previous section, as well as the broader literature on the subject, suggests that the development of a reliable bed load transport formula, valid for a wide range of gravel-bed conditions remains an elusive goal. Though the physical principles governing bed load transport phenomena are universal, the variability exhibited by natural streams is daunting. Furthermore, these physical principles are inadequately represented in the rather simplistic formulations that have been pursued given the complexity of the problem. As a result we end up with multi-valued bed load functions of a given variable. Predictions within an order of magnitude of measured values have been accepted as satisfactory results. Emmett and Wolman (2001), both researchers who have devoted a lifetime of inquiry on the subject, state: ‘‘Given this variety and plethora of controls of channel form, it is not surprising that no all-encompassing relationship between morphology and transport has been

P. Diplas, H. Shaheen

298 1.E+02

Nahal Yatir

1.E+01 1.E+00

Proffitt Initial

1.E-01

Proffitt Final

1.E-02

Jacoby River

W*

Sagehen Creek

1.E-03

Proffitt Final Phase Proffitt Initial Phase

Wilcock and Crowe

1.E-04

Wilcock and Crowe Oak Creek

1.E-05

Oak Creek

Nahal Yatir

1.E-06

Sagehen Creek Jacoby River

1.E-07 1.E-03

1.E-02

τ*

1.E-01

1.E+00

Figure 11.2. Dimensionless bed load transport rate, W*, vs. Shields stress, t*, for several ﬁeld and laboratory datasets.

constructed.’’ Later on though they add that absence of a universal bed load transport equation does not imply ‘‘uniqueness and disorder.’’ Similarly, based on the examination of a large number of bed load measurements in steep, coarse-grained channels in Colorado and Wyoming, Ryan et al. (2005) conclude: ‘‘while there were many similarities in observed patterns of bed load transport at the 19 studied sites, each had its own ‘bed load signal’ in that the rate and size of materials transported largely reﬂected the nature of ﬂow and sediment particular to that system.’’ One way to improve the predictability of a certain equation is to calibrate it by using bed load data from the stream of interest, and thus provide a site-speciﬁc adjustment (e.g., Bakke et al., 1999; Barry et al., 2004). Several bed load datasets have been plotted in Fig. 11.2, in terms of a dimensionless bed load transport parameter, W*, and the Shields stress, t*. They include ﬁeld data from Oak Creek (Milhous, 1973), Jacoby River (Lisle, 1989), Sagehen Creek (Andrews, 1994), Nahal Yatir (Reid et al., 1995) and experimental data obtained by Profﬁtt (1980) (initial and ﬁnal conditions) and Wilcock and Crowe (2003) for the sediment mixture containing 6% sand. The terms are deﬁned as qB t1:5

(11.1)

t0 rgRDm

(11.2)

W ¼

t ¼

Bed load transport and streambed structure in gravel streams qB ¼

299

qB Dm ðRgDm Þ0:5

(11.3)

t0 ¼ rgHS

(11.4)

where qB is the volumetric bed load transport rate, t0 the boundary shear stress, H the ﬂow depth, R ¼ (rrs)/r, r is the ﬂuid density, rs the sediment density, g the acceleration of gravity, Dm the mode size of the surface material and S the channel bed slope. On average, the data follow a trend that has been expressed many times before through the following power formula W ¼ at

m

(11.5)

Equation (11.5) is ﬁtted to each dataset separately, and the corresponding exponents and coefﬁcients are shown in Table 11.1. By representing the range of t* values for each dataset by the corresponding mean value, tmean , the usual reduction in the exponent m with increasing tmean is observed. These results are plotted in Fig. 11.3, to obtain the following expression m ¼ 0:05t1:39 mean

(11.6)

2

with r ¼ 0.94. This result is consistent with the expectation that m will approach zero for high tmean values. Extensive ﬁeld and laboratory observations indicate that some of the consistent trends exhibited by gravel streams include the increase in the degree of coarseness of the bed load material with shear stress and the concomitant improvement in the mobility of the coarser particles, approaching a condition of equal mobility for a shear stress values several times that of the corresponding critical value (e.g., Parker et al., 1982; Diplas, 1987; Powell et al., 2001). Such information is important in determining downstream ﬁning and possible evolution of the armor layer with changing shear stress. To capture these trends, it is necessary to analyze the data in terms of fractional transport rates. This approach is followed here for a small subset of the data shown in Fig. 11.2, namely the Nahal Yatir and the Profﬁtt data. Profﬁtt (1980) conducted experiments in a non-feeding, non-recirculating sediment ﬂume to study the development of an armor layer in the presence of poorly sorted bed material. He used four different sediment mixtures and carried out four experiments with each mixture. He identiﬁed three phases during each experiment. The initial Table 11.1.

Regression results using equation (11.5) for seven bed load datasets.

Data source

t* range

tmean

Exponential (m)

Coefﬁcient (a)

r2

Nahal Yatir Profﬁtt initial Profﬁtt ﬁnal Jacoby River Sagehen Creek Wilcock and Crowe Oak Creek

0.092–0.39 0.046–0.126 0.033–0.051 0.018–0.081 0.027–0.039 0.015–0.050 0.010–0.042

0.241 0.086 0.042 0.050 0.033 0.032 0.026

0.35 1.54 2.63 4.64 4.83 7.65 7.95

7.59 47.40 48.10 1.6 104 3.4 104 1.0 109 3.4 109

0.91 0.70 0.20 0.69 0.26 0.85 0.89

P. Diplas, H. Shaheen

300 9 8 7

m

6 5 4 3 2 1 0 0

0.05

0.1

0.15

0.2

0.25

0.3

τ*

mean

Figure 11.3. Variation of the exponent m for the datasets shown in Fig. 11.2.

phase lasted for about 1 h after the commencement of the experiment and was characterized by intense and relatively constant bed load transport rate in the absence of an armor layer. The ﬁnal phase was characterized by the highest degree of bed surface armoring, which was distinct for each of the 16 experiments, a condition reached after 20–95 h of run time and a bed load transport rate of 2.5% or lower of that measured during the initial phase of the corresponding experiment. During the intermediate phase, the channel bottom and bed load transitioned between the initial and ﬁnal phases. The data collected by Profﬁtt during the initial and ﬁnal phases of his experiments are analyzed here separately in terms of fractional transport rates. The original sediment mixture, used as the composition of the bed surface layer for the initial phase, and that of the ﬁnal armor layer, used for the ﬁnal phase, are divided into 10 size ranges starting with Di ¼ 0.70 mm and ending with Di ¼ 15.55 mm (Table 11.2), where Di is the geometric mean diameter of the ith grain size range. Material ﬁner than 0.5 mm has been assumed to be transported, at least in part, in suspension and thus has been excluded from further consideration. This material comprised between 0.2 and 1.7% of the total sediment transported during a run. Equations (11.1)–(11.3) and (11.5) are adjusted for fractional transport as follows qBi t1:5 i

(11.7)

t0 rgRDi

(11.8)

W i ¼ ti ¼

qBi ¼

qBi f i Di ðRgDi Þ0:5

i W i ¼ ai tm i

(11.9)

(11.10)

Bed load transport and streambed structure in gravel streams Table 11.2. data.

301

Regression results of W i ¼ ai tmi for initial and ﬁnal phases of Profﬁtt’s bed load i

Size range (mm)

0.60–0.85 0.85–1.20 1.20–1.68 1.68–2.41 2.41–3.35 3.35–4.76 4.76–6.35 6.35–9.52 9.52–12.7 12.7–19.0 Average/total

Di (mm)

0.72 1.01 1.42 2.01 2.84 3.99 5.50 7.78 11.00 15.55

Final phase data

Initial phase data

mi

aI

r2

%fiav

mi

ai

r2

%fiav

1.78 1.28 1.35 1.67 1.86 2.28 2.65 2.83 3.81 5.05 2.46

0.09 0.09 0.18 0.61 1.63 7.81 33.05 39.17 1.3 103 5.5 104

0.20 0.14 0.22 0.38 0.45 0.51 0.48 0.55 0.39 0.38 0.37

2.07 4.29 4.82 6.58 5.23 13.94 17.64 7.89 12.68 16.27 91.41

2.21 1.97 2.02 1.96 1.93 1.90 2.06 2.14 3.87 3.69 2.38

6.65 14.66 33.72 64.83 121.86 208.74 528.51 975.28 5.14 105 2.10 105

0.57 0.66 0.75 0.77 0.78 0.84 0.84 0.77 0.55 0.32 0.69

3.91 7.13 10.01 13.67 11.05 16.00 16.50 5.41 4.54 4.61 92.83

where fi is the percent of the reference material, original mixture or armor layer, contained within Di; qBi is the volumetric bed load transport rate of size fraction represented by the diameter Di. Plots of W i vs. ti , for both the initial and the ﬁnal phases are shown in Fig. 11.4. The exponents mi and coefﬁcients ai, obtained by ﬁtting equation (11.10) to the bed load data of each size range, are shown in Table 11.2. For the initial phase data analysis, the mi values for the ﬁrst eight size ranges do not vary much and do not show any speciﬁc trend. It is reasonable to assume that they are transported under equal mobility conditions. No information is lost by combining them into a single size range. The same is true for the last two size ranges, which are similarly combined into one. The ﬁnal phase exponents, except for Di ¼ 0.72, show consistent but very gradual changes, increasing with Di. This suggests that the difference in mobility of consecutive size ranges is very modest and it is not necessary to maintain 10 ranges. These therefore are reduced to 4, as shown in Table 11.4. Equation (11.10) is ﬁtted once again to the fractional bed load data of the initial and ﬁnal phases, based on the fewer size ranges. The corresponding mi and ai values are included in Tables 11.3 and 11.4, for the initial and ﬁnal phases, respectively. In both cases, the mi values exhibit consistent behavior, increasing with size range. More speciﬁcally, these trends are described by regression equations (11.11) and (11.12) for the initial and ﬁnal phases respectively

Di mi ¼ 2:08 D50 mi ¼ 2:69

Di D50

0:4 (11.11)

0:4 (11.12)

P. Diplas, H. Shaheen

302 1.E+01 7.78

11

5.5

3.99

2.84

2.0

1.42

1.0

0.72mm

1.E+00 15.6 Intialph asebedload data 7.78

1.E-01

5.5

3.99

2.84

2.0

W*r= 0.35 0.72mm

1.42 1.0

11

W*i

1.E-02 15.6

W*r=0.0025

1.E-03 Final phase bedload data

1.E-04 Dimm

1.E-05 1.E-06 1.E-02

1.E-01 τ*

1.E+00

Figure 11.4. Plot of W i vs. ti for Profﬁtt’s fractional bed load transport data during initial and ﬁnal phases. Ten size ranges for each case are considered.

Table 11.3. Regression results using equations (10) and (13) for Profﬁtt’s initial phase data using two grain size ranges. Size range (mm)

Di (mm) (1)

mi (2)

ai (3)

r2 (4)

m0i (5)

0.85–9.52 9.52–19.0

2.85 13.45

1.95 3.58

117.56 2.37 105

0.82 0.52

2.08 2.08

Table 11.4. Regression results using equations (10) and (13) for Profﬁtt’s ﬁnal phase data using four grain size ranges. Size range (mm)

Di (mm) (1)

mi (2)

ai (3)

r2 (4)

m0i (5)

0.85–1.68 1.68–3.35 3.35–9.52 9.52–19.0

1.20 2.37 5.65 13.45

1.29 1.77 2.44 3.41

0.13 0.98 17.65 315.55

0.18 0.43 0.45 0.40

2.64 2.75 2.69 2.66

Following Diplas (1987), a new independent variable is used, that takes into account 0:4equations (11.11) and (11.12), and the data are plotted in terms of W i vs. ðD =D Þ ti i 50 (Fig. 11.5). The resulting regression equations are described by ðDi =D50 Þ0:4 m0i

W i bi ½ti

(11.13)

The exponents m0i , shown in Tables 11.3 and 11.4 for the two phases, exhibit very small variability within each phase, suggesting that the new variable has rendered the

Bed load transport and streambed structure in gravel streams

303

10 2.08

W*i=W *r (x) 2 r =0.76 W*r = 0.0025

1

0.1

2.67

W*i=W*r (x) 2 r =0.81 W*r = 0.0025

W*i

0.01

0.001

0.0001 Final Phase (Proffitt Data) Initial Phase (Proffitt Data)

0.00001

0.000001 0.1

1

10

100

x = (T*i/T*r)^(D i/D50)^0 .4 ðDi =D50 Þ^ 0:40

Figure 11.5. W i vs. t Profﬁtt’s bed load data.

for the initial (two size ranges) and ﬁnal (four size ranges) phases of

curves ﬁtting the fractional data geometrically similar. This in turn, indicates that the new variable accounts for the differences in mobility among the particles in the various size ranges. The bed load transport rates measured in Nahal Yatir, represent some of the highest rates ever reported in the literature (Laronne et al., 1994). For the very high shear stress values considered here (at least 4.5 times the critical Shields stress), they reﬂect conditions of equal mobility with respect to the surface material. As such, there is no need to consider fractional transport analysis for this stream. The corresponding regression equation for this case is W ¼ 7:59t

0:35

(11.14)

The analysis of this limited number of datasets suggests that as the Shields stress value increases (see Table 11.1) fewer sediment fractions are necessary to capture the trends exhibited by the bed load transport data. Four and two fractions for the ﬁnal and initial phases of Profﬁtt’s experiments respectively, while a fractional approach was not deemed necessary for the ﬁeld data from Nahal Yatir. This is consistent with a large number of observations indicating that the equal mobility condition becomes a closer representation of reality with increasing Shields stress values. Because of the improved similarity obtained with the new variable, all the data analyzed here can be collapsed on a single curve by considering the reference shear stress value, tri, that corresponds to W i ¼ W r ¼ 0:0025 for each size fraction; where W r ¼ 0:0025 represents a very low transport rate condition (Parker et al., 1982) (Fig. 11.6). In this

P. Diplas, H. Shaheen

304 1.E+04 1.E+03 1.E+02 1.E+01

W*i

1.E+00 1.E-01 1.E-02 1.E-03

Final Phase (Proffitt)

1.E-04

Initial Phase (Proffitt)

1.E-05

Nahal Yatir

1.E-06 1.E-01 1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E+05 1.E+06 1.E+07 1.E+08 1.E+09 1.E+10 (Di/D50)^0.4'

(τ*i /τ*r)

Figure 11.6. Plot of the modiﬁed similarity collapse for Profﬁtt’s (initial and ﬁnal phases) and Nahal Yatir data.

case, the tri values for Profﬁtt’s data are obtained from equation (11.13) and for Nahal Yatir from equation (11.14). Although additional data for intermediate values of bed load transport and shear stress are required to verify the trends obtained here, Fig. 11.6 suggests that fractional transport rates, like the total transport rates shown in Fig. 11.2, follow consistent trends, when proper transformations are employed that account for the gradual adjustments exhibited by the data, from selective transport dominated events to conditions where equal mobility prevails.

5.

Streambed structure

The conventional thinking about the evolution of the armor layer has been that it achieves its coarsest state under sediment-starved conditions, which also represent the threshold of motion condition. As the shear stress increases, the armor layer weakens and at reasonably high shear stresses it disappears, rendering the bed material uniform in the vertical direction (e.g., Parker and Klingeman, 1982; Dietrich et al., 1989; Parker, 1990). This thinking is well supported by ﬁeld data of streams operating at near threshold and very high shear stresses (e.g., for ephemeral streams). There is a paucity of data though at active bed conditions, given the difﬁculty of collecting such data. Andrews and Erman (1986) provide one of the few, if not the only, widely quoted measurements of an armor layer during a low ﬂow and a ﬂood event in Sagehen Creek, a perennial stream in the Sierra Nevada of California. The median size of the armor layer, D50, at low ﬂow was 58 mm and during the peak of

Bed load transport and streambed structure in gravel streams

305

the ﬂood it was 46 mm. The corresponding value reported for the subsurface was 30 mm. Proponents and opponents of the conventional thinking interpret this result differently. Similarly, results from sediment feed ﬂumes have been used to support the conventional thinking (e.g., Kuhnle and Southard, 1988) but they have been countered by evidence from experiments in sediment recirulating ﬂumes that show insensitivity of the armor layer composition to varying shear stresses (e.g., Wilcock and DeTemple, 2005). However, it is well known that, in general, neither ﬂume type accurately represents the complex ﬁeld conditions. In the ﬁrst ﬂume type, the constant feed rate, in amount and composition, dictates, to a great extent the outcome. In the second ﬂume type, the initial conditions (the composition of the bed material placed in the ﬂume) dictate the outcome. Conservation of mass (bed material) requires that after the ﬁner particles have inﬁltrated below the surface layer, there is very little room for further adjustments in the make up of the armor layer except, possibly, at very high shear stresses when multiple layers (at least three) of bed material are getting entrained by the ﬂow. There are cases, though, when one ﬂume type is better suited to model a speciﬁc phenomenon. A case in point is the development of a static armor layer under starved sediment conditions, e.g., simulating the response of a streambed due to the construction of a dam. For the feeding case, this will be approached by cutting off the sediment input at the upstream end, while the transported sediment is allowed to escape at the downstream end of the ﬂume. For the recirculating case, this phenomenon can be modeled by gradual reduction in water discharge, or channel bed slope, or both in an effort to approach conditions at or below threshold. The former is closer to reality in this case. In general though, a long ﬂume having a test section at its downstream end, without sediment feed, will provide conditions that are closer to a natural setting. Cost and space considerations, however, preclude the wide availability of such facilities. The US Geological Survey and US Forest Service data (King et al., 2004) employed by Whiting and King (2003) and plotted in Fig. 11.1 indicate that an increase in bed load coarseness is associated with a weaker armor layer. Although it is difﬁcult to separate the inﬂuence of sediment supply on this trend, the large range of D50/D50s values (from 1.9 to 7.2) is worth pointing out. Given the relative constancy of the subsurface material composition, this result hardly points to a degree of armoring that remains constant regardless of the ﬂow and other conditions. Finally, the results from a large number of ephemeral streams, operating under very high shear stresses and sediment supply, and the absence of an armor layer thereof, further support the notion that the armor layer weakens with ﬂood stage, assuming that a critical value is exceeded. In many perennial streams, this value will represent an infrequent ﬂood event.

6.

Conclusions

The signiﬁcant contributions, in number and quality that have been made recently in gravel-bed river mechanics, have provided some new insights, reinforced prior results and emphasized the complexity of these stream types. While the development of the ‘‘general equation’’ of bed load transport might not be imminent, sufﬁcient information

306

P. Diplas, H. Shaheen

exists to identify well-deﬁned trends, even at the fractional transport level. Guidance also exists on how to deal with engineering aspects of the problem. The subject is approached in a very systematic way, with both extensive ﬁeldwork and laboratory experimentation. Various issues are tackled at different scales, from the individual grain to the entire watershed. More data, and of better quality, continue to be gathered. The subject is ripe for some signiﬁcant developments.

Acknowledgment This work was supported by the National Science Foundation (EAR-0439663). References Ackers, P., White, W.R., 1973. Sediment transport: a new approach and analysis. J. Hydraul. Eng. ASCE 99 (11), 2041–2060. Almedeij, J.H., Diplas, P., 2003. Bedload transport in gravel-bed streams with unimodal sediment. J. Hydraul. Eng. ASCE 129 (11), 896–904. Andrews, E.D., 1994. Marginal bed load transport in a gravel bed stream, Sagehen Creek, California. Water Resour. Res. 30 (7), 2241–2250. Andrews, E.D., 2000. Bed material transport in the Virgin River, Utah. Water Resour. Res. 36 (2), 585–596. Andrews, E.D., Erman, D.C., 1986. Persistence in the size distribution of surﬁcial bed material during an extreme snowmelt ﬂood. Water Resour. Res. 22 (2), 191–197. Ashida, K., Michiue, M., 1972. Study on hydraulic resistance and bedload transport rate in alluvial streams. Trans. Japan Soc. Civil Eng. 206, 59–69, (in Japanese). Bagnold, R.A., 1966. An approach to the sediment transport problem from general physics. U.S. Geological Survey Professional Paper 422-I, 37pp. Bakke, P.D., Baskedas, P.O., Dawdy, D.R., Klingeman, P.C., 1999. Calibrated Parker-Klingeman model for gravel transport. J. Hydraul. Eng. ASCE 125 (6), 657–660. Barry, J.J., Bufﬁngton, J.M., King, J.G., 2004. A general power equation for predicting bed load transport rates in gravel bed rivers. Water Resour. Res. 40, W10401, 10.1029/2004WR003190. Barry, J.J., Bufﬁngton, J.M., King, J.G., 2005. Reply to comment on ‘‘A general power equation for predicting bed load transport rates in gravel bed rivers’’. Water Resour. Res. 41, W07016, 10.1029/ 2005WR004172. Bufﬁngton, J.M., Montgomery, D.R., 1997. A systematic analysis of eight decades of incipient motion studies, with special reference to gravel-bedded rivers. Water Resour. Res. 33, 1993–2029. Bufﬁngton, J.M., Montgomery, D.R., 1999. Effects of sediment supply on surface textures of gravel-bed rivers. Water Resour. Res. 35 (11), 3523–3530. Church, M., Hassan, M.A., 2002. Mobility of bed material in Harris Creek. Water Resour. Res. 38 (11), 1237, 10.1029/2001WR000753. Cohen, H., Laronne, J.B., 2005. High rates of sediment transport by ﬂashﬂoods in the Southern Judean Desert, Israel. Hydrol. Process. 19, 1687–1702. Dietrich, W.E., Kirchner, J.W., Ikeda, H., Iseya, F., 1989. Sediment supply and the development of the coarse surface layer in gravel-bedded rivers. Nature 340, 215–217. Diplas, P., 1987. Bedload transport in gravel-bed streams. J. Hydraul. Eng. ASCE 113 (3), 277–292. DuBoys, P., 1879. ‘‘Le Rohne et les Rivieres a Lit Affouillable,’’ Annales des Pontes et Chausses Series 5, Vol. 18, pp. 141–195. Egiazaroff, I.V., 1965. Calculation of non-uniform sediment concentrations. J. Hydraul. Div. ASCE 91 (4), 225–248. Einstein, H.A., 1950. The bed load function for sediment transportation in open channels. Technical Bulletin 1026, U.S. Department of Agriculture, Soil Conservation Service, Washington, DC.

Bed load transport and streambed structure in gravel streams

307

Einstein, H.A., El-Samni, E.S.A., 1949. Hydrodynamic forces on a rough wall. Rev. Mod. Phys. 21, 520–524. Emmett, W.W., Wolman, M.G., 2001. Effective discharge and gravel-bed rivers. Earth Surf. Process. Landf. 26, 1369–1380. Gilbert, G.K., 1914. The transportation of debris by running water. U.S. Geological Survey Professional Paper 86, Washington, DC. Gomi, T., Sidle, R.C., 2003. Bed load transport in managed steep-gradient headwater streams of southeastern Alaska. Water Resour. Res. 39 (12), 1336, 10.1029/2003WR002440. Harrison, A.S., 1950. Report on special investigations of bed sediment segregation in a degrading bed. Report Series No. 33, University of California, Institute of Engineering Research, Issue No. 1, Berkeley, CA, September. Haschenburger, J.K., Wilcock, P.R., 2003. Partial transport in a natural gravel bed channel. Water Resour. Res. 39 (1), 1020, 10.1029/2002WR001532. Helley, E.J., Smith, W., 1971. Development and calibration of a pressure-difference bedload sampler. U.S. Geological Survey Open-ﬁle Report, 18. Jackson, W.L., Beschta, R.L., 1982. A model of two-phase bedload transport in an Oregon Coast Range stream. Earth Surf. Process. Landf. 7, 517–527. Johnston, W.A., 1922. Imbricated structures in river-gravels. Am. J. Sci. IV (24), 387–390. Kellerhals, R., 1967. Stable channels with gravel-paved beds. Waterways Harbors Div. ASCE 93 (WW1), 63–84. Kellerhals, R., Bray, D.I., 1971. Sampling procedures for coarse ﬂuvial sediments. J. Hydraul. Div. ASCE 97, 1165–1180. King, J.G., Emmett, W.W., Whiting, P.J., Kenworthy, R.P., Barry, J.J., 2004. Sediment transport data and related information for selected coarse-bed streams and rivers in Idaho. U.S. Forest Service Technical Report RMRS-GTR-131, 26pp. Kuhnle, R.A., Southard, J.B., 1988. Bed load transport ﬂuctuations in a gravel bed laboratory channel. Water Resour. Res 24 (2), 247–260. Lane, E.W., Carlson, E.J., 1953. Some Factors Affecting the Stability of Canals Constructed in Coarse Granular Materials. International Association for Hydraulic Research, Minneapolis, MN. Laronne, J.B., Reid, I., 1993. Very high rates of bedload sediment transport by ephemeral desert streams. Nature 366, 148–150. Laronne, J.B., Reid, I., Yitsak, Y., Frostick, L.E., 1994. The non-layering of gravel streambeds under ephemeral ﬂood regimes. J. Hydrol. 159, 353–363. Lisle, T.E., 1989. Sediment transport and resulting deposition in spawning gravels, north Coastal California. Water Resour. Res. 25 (6), 1303–1319. Lisle, T.E., 1995. Particle size variations between bed load and bed material in natural gravel bed channels. Water Resour. Res. 31 (4), 1107–1118. Lisle, T.E., Nelson, M.N., Pitlick, J., et al., 2000. Variability of bed mobility in natural, gravel-bed channels and adjustments to sediment load at local and reach scales. Water Resour. Res. 36 (12), 3743–4755. Martin-Vide, J.P., Ninerola, D., Bateman, A., et al., 1999. Runoff and sediment transport in a torrential ephemeral stream of the Mediterranean coast. J. Hydrol. 225, 118–129. Meyer-Peter, E., Muller, R., 1948. Formulas for bed-load transport. Proceedings, 2nd Congress International Association for Hydraulic Research, Stockholm, Sweden, pp. 39–64. Milhous, R.T., 1973. Sediment transport in a gravel-bottomed stream. Ph.D. Thesis, Oregon State University, Corvallis, Oregon. Mueller, E.R., Pitlick, J., Nelson, J.M., 2005. Variation in the reference Shields stress for bed load transport in gravel-bed streams and rivers. Water Resour. Res. 41, W04006, 10.1029/2004WR003692. Neil, C.R., 1968. A reexamination of the beginning of movement for coarse granular bed materials. Report INT 68, Hydraulics Research Station, Wallingford, England. Pantelopulos, J., 1955. Note sur la granulometrie de charriage et la loi du debit solide par charriage de fond d’un me´lange de materiaux. 6th Congress International Association for Hydraulic Research, Hague, report D-10. Pantelopulos, J., 1957. Etude experimentale du movement par charriage de fond d’un me´lange de materiaux. 7th Congress International Association for Hydraulic Research, Lisbon, Portugal, report D-30.

308

P. Diplas, H. Shaheen

Parker, G., 1990. Surface-based bedload transport relation for gravel rivers. J. Hydraul. Res. 28 (4), 417–436. Parker, G., Klingeman, P.C., 1982. On why gravel bed streams are paved. Water Resour. Res. 18, 1409–1423. Parker, G., Klingeman, P.C., McLean, D.L., 1982. Bedload and size distribution in paved gravel-bed streams. J. Hydraul. Div. ASCE 108 (4), 544–571. Parker, G., Wilcock, P.R., 1993. Sediment feed and recirculating ﬂumes: a fundamental difference. J. Hydraul. Eng. ASCE 119 (11), 1192–1204. Powell, D.M., Reid, I., Laronne, J.B., 2001. Evolution of bed load grain size distribution with increasing ﬂow strength and the effect of ﬂow duration on the caliber of bed load sediment yield in ephemeral gravel bed rivers. Water Resour. Res. 37 (5), 1463–1474. Profﬁtt, G.T., 1980. Selective transport and armouring of non-uniform alluvial sediments. Report no. 80/ 22, Department of Civil Engineering, University of Canterbury, Christchurch, NZ. Reid, I., Frostick, L.E., Layman, J.T., 1985. The incidence and nature of bedload transport during ﬂood ﬂows in coarse-grained alluvial channels. Earth Surf. Process. Landf. 13, 33–44. Reid, I., Laronne, J.B., 1995. Bed load sediment transport in an ephemeral stream and a comparison with seasonal and perennial counterparts. Water Resour. Res. 31 (3), 773–781. Reid, I., Laronne, J.B., Powell, D.M., 1995. The Nahal Yatir bedload database sediment dynamics in a gravel-bed ephemeral stream. Earth Surf. Process. Landf. 20, 845–857. Reid, I., Laronne, J.B., Powell, D.M., 1998. Flash-ﬂood and bedload dynamics of desert gravel-bed streams. Hydrol. Process. 12, 543–557. Reid, I., Powell, D.M., Laronne, J.B., 1996. Prediction of bed-load transport by desert ﬂash ﬂoods. J. Hydraul. Eng. ASCE 122 (3), 170–172. Richter, I.A., 1998. The notebooks of Leonardo Da Vinci. Oxford, 417pp. Rickenmann, D., 2001. Comparison of bed load transport in torrents and gravel bed streams. Water Resour. Res. 37 (12), 3295–3305. Ryan, S.E., Porth, L.S., Troendle, C.A., 2005. Coarse sediment transport in mountain streams in Colorado and Wyoming, USA. Earth Surf. Process. Landf. 30, 269–278. Schoklitsch, A., 1962. Handbuch des Wasserbaus, 3rd ed., Springer-Verlag, New York. Shields, I.A., 1936. Applications of similarity principles and turbulence research to bed-load movement. (in German). Mitt. Der Preuss. Versuchsanstalt fur Wasserbau und Schiffbau, Berlin, 1936. English translation by W.P Ott and J.C. van Uchelen, California Institute of Technology, Hydrodyn. Lab. Publ. 167, Pasadena, CA. Sundborg, A., 1956. The River Klaralven, a study of ﬂuvial processes, Geograﬁska Annaler 127–316. Sternberg, H., 1875. Untersuchungen uber Langen und Querproﬁl deschiebefuhrende Flusse. Z. Bauwesen 25, 483–506. Vanoni, V.A. (Ed.), 1975. Sedimentation Engineering. ASCE Manuals and Reports on Engineering Practice, No. 54, New York, NY, 745pp. Whiting, P.J., King, J.G., 2003. Surface particle sizes on armoured gravel streambeds: effects of supply and hydraulics. Earth Surf. Process. Landf. 28, 1459–1471. Wilcock, P.R., 2001. The ﬂow, the bed, and the transport: interaction in ﬂume and ﬁeld. Proceedings, 5th Gravel-Bed Rivers Workshop, Christchurch, NZ. Wilcock, P.R., Crowe, J.C., 2003. Surface-based transport model for mixed-size sediment. J. Hydraul. Eng. ASCE 129 (2), 120–128. Wilcock, P.R., DeTemple, B.T., 2005. Persistence of armor layer in gravel-bed streams. Geophys. Res. Lett. 32, L08402, 10.1029/2004GL021772. Wilcock, P.R., Kenworthy, S.T., Crowe, J.C., 2001. Experimental study of the transport of mixed sand and gravel. Water Resour. Res. 37 (12), 3349–3358. Wilcock, P.R., McArdell, B.W., 1993. Surface-based fractional transport rates: mobilization thresholds and partial transport of a sand-gravel sediment. Water Resour. Res. 29, 1297–1312. Wolman, M.G., 1953. A method of sampling coarse river bed material. Transactions of American Geophysical Union Vol. 35, No. 6, December. Yatsu, E., 1957. On the discontinuity of grainsize frequency distribution of ﬂuvial deposits and its geomorphological signiﬁcance. Proceedings, IGU Regional Conference, Japan, 1957, pp. 224–237.

Bed load transport and streambed structure in gravel streams

309

Discussion by Lynne E. Frostick In your paper you speculate that the observed differences in bed structure between perennial and ephemeral streams (Reid and Laronne, 1995) may be linked with the bed ﬂuidizing in some way during the peak ﬂows of ﬂash ﬂood events, bringing coarser particles to the surface and preventing the bed becoming organized in the manner frequently reported from the better studied perennial counterparts. Inevitably there are no ﬁeld observations to support or contradict this assertion since ﬂash ﬂoods are notoriously difﬁcult to predict and instrument (Reid et al., 1998). However, some laboratory experiments carried out at the University of Hull give some hints that ﬂuidization may be a product of ﬂows which generate shear well in excess of critical for the bed material, but only in the presence of mixed grain sizes. These experiments used digital video recordings to monitor entrainment in unimodal gravels (8 mm) and sand–gravel mixtures (with 0.9 mm sand added). Image analysis showed that particles in the unimodal gravels entrained singly and in a continual manner, with the entrainment process only affecting the surface layer of grains. In the sand–gravel mixtures, entrainment was patchy and sporadic with extensive entrainment switching from one area of the bed to another. In addition, the framework of the gravels was observed to dilate beneath the entraining particles to a depth of more than six particle diameters (Fig. 11.7) and analysis of particle trajectories at entrainment suggest lift is the most important instigator of entrainment (Allan and Frostick, 1999). Both of these observations may link to opening up of the bed structure and upward movements of ﬂuid through the bed both of which are essential precursors to ﬂuidization. The lack of such changes during entrainment in unimodal gravels suggest some limitations on the process, which link to bed character, not ﬂow alone.

Figure 11.7. Differenced image for sand–gravel mixtures. Flow is from left to right. White spaces are the pebbles and black areas denote movement between successive video images. Note the penetration of movement well beneath the surface in the right half of the image.

310

P. Diplas, H. Shaheen

References Allan, A., Frostick, L.E., 1999. Framework dilation, winnowing, and matrix particle size: The behaviour of some sand-gravel mixtures in a laboratory ﬂume. J. Sediment. Res. 69, 21–26. Reid, I., Laronne, J.B., 1995. Bedload sediment transport in an ephemeral stream and a comparison with seasonal and perennial counterparts. Water Resour. Res. 31, 773–781. Reid, I., Laronne, J.B., Powell, D.M., 1998. Flash-ﬂood and bedload dynamics of desert gravel-treams. Hydrol. Process. 12, 543–557.

Discussion by Gordon E. Grant The concept of timescale is perhaps underappreciated in thinking about how structure evolves in river systems and may perhaps underlie the distinction that the author draws between ephemeral and perennial rivers. Is there a characteristic timescale of sediment transport required to produce the armoring that one typically sees in perennial but is often absent in ephemeral channels? In other words, could the short duration of ﬂood ﬂows in ephemeral streams be insufﬁcient to establish an armored bed?

Discussion by Murray Hicks and John Laronne Panos drew attention to differences in what amounts to bedload transport efﬁciency between ephemeral rivers (high efﬁciency) and perennial rivers (lower efﬁciency), noting how sustained lower ﬂows tend to coarsen the bed surface texture in perennial rivers, thereby limiting the availability of entrainable bed material. We note that perennial braided rivers often show a continuum of such ephemeral and perennial characteristics. On braided riverbeds, perennial low ﬂows occupy typically well-armored channels that cover only a small fraction of the braidplain area. While some of the dry riverbed consists of relict channels that also tend to be armored, much is formed from gravel bars and sheets that are only active ephemerally during ﬂoods. In our experiences from braided rivers that rise and fall quickly, these surfaces typically show the poorly armored, minimally reworked surface textures that Panos notes are characteristic of ﬂashy ephemeral single-thread rivers. Indeed, on braided riverbeds the rapid rate of change of width with discharge ampliﬁes the temporal rate of width change (i.e., dW/dt ¼ dW/dQ dQ/dt), thus on recessions bars are abandoned quickly as ﬂows converge and cut into a few channels, and there is little opportunity for reworking. The implication is that bedload transport efﬁciency (equating to bedload availability) over a braidplain during ﬂoods also ranges over a continuum. This should certainly be considered when computing bedload transport rates over braidplains, and surface-based formulations such as those of Parker (1990) and Wilcock and Crowe (2003) would seem to be the best ways to capture this, provided there exists spatially distributed information on bed surface texture.

310

P. Diplas, H. Shaheen

References Allan, A., Frostick, L.E., 1999. Framework dilation, winnowing, and matrix particle size: The behaviour of some sand-gravel mixtures in a laboratory ﬂume. J. Sediment. Res. 69, 21–26. Reid, I., Laronne, J.B., 1995. Bedload sediment transport in an ephemeral stream and a comparison with seasonal and perennial counterparts. Water Resour. Res. 31, 773–781. Reid, I., Laronne, J.B., Powell, D.M., 1998. Flash-ﬂood and bedload dynamics of desert gravel-treams. Hydrol. Process. 12, 543–557.

Discussion by Gordon E. Grant The concept of timescale is perhaps underappreciated in thinking about how structure evolves in river systems and may perhaps underlie the distinction that the author draws between ephemeral and perennial rivers. Is there a characteristic timescale of sediment transport required to produce the armoring that one typically sees in perennial but is often absent in ephemeral channels? In other words, could the short duration of ﬂood ﬂows in ephemeral streams be insufﬁcient to establish an armored bed?

Discussion by Murray Hicks and John Laronne Panos drew attention to differences in what amounts to bedload transport efﬁciency between ephemeral rivers (high efﬁciency) and perennial rivers (lower efﬁciency), noting how sustained lower ﬂows tend to coarsen the bed surface texture in perennial rivers, thereby limiting the availability of entrainable bed material. We note that perennial braided rivers often show a continuum of such ephemeral and perennial characteristics. On braided riverbeds, perennial low ﬂows occupy typically well-armored channels that cover only a small fraction of the braidplain area. While some of the dry riverbed consists of relict channels that also tend to be armored, much is formed from gravel bars and sheets that are only active ephemerally during ﬂoods. In our experiences from braided rivers that rise and fall quickly, these surfaces typically show the poorly armored, minimally reworked surface textures that Panos notes are characteristic of ﬂashy ephemeral single-thread rivers. Indeed, on braided riverbeds the rapid rate of change of width with discharge ampliﬁes the temporal rate of width change (i.e., dW/dt ¼ dW/dQ dQ/dt), thus on recessions bars are abandoned quickly as ﬂows converge and cut into a few channels, and there is little opportunity for reworking. The implication is that bedload transport efﬁciency (equating to bedload availability) over a braidplain during ﬂoods also ranges over a continuum. This should certainly be considered when computing bedload transport rates over braidplains, and surface-based formulations such as those of Parker (1990) and Wilcock and Crowe (2003) would seem to be the best ways to capture this, provided there exists spatially distributed information on bed surface texture.

310

P. Diplas, H. Shaheen

References Allan, A., Frostick, L.E., 1999. Framework dilation, winnowing, and matrix particle size: The behaviour of some sand-gravel mixtures in a laboratory ﬂume. J. Sediment. Res. 69, 21–26. Reid, I., Laronne, J.B., 1995. Bedload sediment transport in an ephemeral stream and a comparison with seasonal and perennial counterparts. Water Resour. Res. 31, 773–781. Reid, I., Laronne, J.B., Powell, D.M., 1998. Flash-ﬂood and bedload dynamics of desert gravel-treams. Hydrol. Process. 12, 543–557.

Discussion by Gordon E. Grant The concept of timescale is perhaps underappreciated in thinking about how structure evolves in river systems and may perhaps underlie the distinction that the author draws between ephemeral and perennial rivers. Is there a characteristic timescale of sediment transport required to produce the armoring that one typically sees in perennial but is often absent in ephemeral channels? In other words, could the short duration of ﬂood ﬂows in ephemeral streams be insufﬁcient to establish an armored bed?

Discussion by Murray Hicks and John Laronne Panos drew attention to differences in what amounts to bedload transport efﬁciency between ephemeral rivers (high efﬁciency) and perennial rivers (lower efﬁciency), noting how sustained lower ﬂows tend to coarsen the bed surface texture in perennial rivers, thereby limiting the availability of entrainable bed material. We note that perennial braided rivers often show a continuum of such ephemeral and perennial characteristics. On braided riverbeds, perennial low ﬂows occupy typically well-armored channels that cover only a small fraction of the braidplain area. While some of the dry riverbed consists of relict channels that also tend to be armored, much is formed from gravel bars and sheets that are only active ephemerally during ﬂoods. In our experiences from braided rivers that rise and fall quickly, these surfaces typically show the poorly armored, minimally reworked surface textures that Panos notes are characteristic of ﬂashy ephemeral single-thread rivers. Indeed, on braided riverbeds the rapid rate of change of width with discharge ampliﬁes the temporal rate of width change (i.e., dW/dt ¼ dW/dQ dQ/dt), thus on recessions bars are abandoned quickly as ﬂows converge and cut into a few channels, and there is little opportunity for reworking. The implication is that bedload transport efﬁciency (equating to bedload availability) over a braidplain during ﬂoods also ranges over a continuum. This should certainly be considered when computing bedload transport rates over braidplains, and surface-based formulations such as those of Parker (1990) and Wilcock and Crowe (2003) would seem to be the best ways to capture this, provided there exists spatially distributed information on bed surface texture.

Bed load transport and streambed structure in gravel streams

311

References Parker, G., 1990. Surface-based bedload transport relation for gravel rivers. J. Hydraul. Res. 28 (4), 417–436. Wilcock, P.R., Crowe, J.C., 2003. Surface-based transport model for mixed-size sediment. J. Hydraul. Eng. 29 (2), 120–128.

Reply by the authors All three discussions deal with various aspects of the differences in the streambed structure exhibited by perennial and ephemeral streams, mainly the presence of an armor layer in the former and its absence in the latter. It is reasonable to consider the different ﬂow/sediment conditions that these two stream types operate under in an effort to develop a cause and effect relationship. Perennial streams are characterized by continuous base ﬂow interrupted by infrequent ﬂoods generating boundary shear stresses modestly higher than the required critical entrainment value based on the median size of the bed material. Ephemeral streams are typically dry except for short periods of time when rare ﬂoods generate very high boundary shear stresses. Furthermore, ephemerals experience an abundance of sediment supply while perennials usually suffer from limited sediment input. It is worth mentioning though that ephemeral streams from different regions encounter considerable variability in ﬂow/sediment conditions and exhibit a range of streambed material stratiﬁcations (e.g., Tooth, 2000; Hassan et al., 2006). The same is true for perennial streams (e.g., Whiting and King, 2003; Hassan et al., 2006). Grant suggests that the different timescales associated with the two stream types might be the driving mechanism resulting in the two distinct streambed structures. While it is difﬁcult to provide direct evidence in support or against this suggestion, indirect evidence exists indicating that ﬂood duration is important in enhancing the development of an armor layer but might not be the critical parameter identifying existence or absence of a coarser surface layer. For example, there are ephemeral streams that experience ﬂooding for sufﬁciently long time, up to a few days, and yet do not exhibit stratiﬁcation in terms of grain size (e.g., Greenbaum and Bergman, 2006). Furthermore, laboratory experiments, operating in sediment feed, sediment starved and recirculation modes, have shown that the bed material and bed load transport rates undergo signiﬁcant changes during the ﬁrst hour of the experiments (Profﬁtt, 1980; Diplas and Parker, 1985). However, in many cases, the starved sediment being the most noticeable ones, the degree of armoring becomes more pronounced with the duration of the experiment (Profﬁtt, 1980). One exception to the above statements is with regard to the duration of the receding portion of a ﬂood in ephemeral streams. A sufﬁciently long receding ﬂow could facilitate the establishment of an armor layer (Hassan et al., 2006). Strong evidence also exists about the importance of sediment supply on the make up of the surface layer (e.g., Whiting and King, 2003; Hassan et al., 2006). During the presentation, the authors suggested that the mechanism responsible for sediment transport during boundary shear stresses is much higher than those

Bed load transport and streambed structure in gravel streams

311

References Parker, G., 1990. Surface-based bedload transport relation for gravel rivers. J. Hydraul. Res. 28 (4), 417–436. Wilcock, P.R., Crowe, J.C., 2003. Surface-based transport model for mixed-size sediment. J. Hydraul. Eng. 29 (2), 120–128.

Reply by the authors All three discussions deal with various aspects of the differences in the streambed structure exhibited by perennial and ephemeral streams, mainly the presence of an armor layer in the former and its absence in the latter. It is reasonable to consider the different ﬂow/sediment conditions that these two stream types operate under in an effort to develop a cause and effect relationship. Perennial streams are characterized by continuous base ﬂow interrupted by infrequent ﬂoods generating boundary shear stresses modestly higher than the required critical entrainment value based on the median size of the bed material. Ephemeral streams are typically dry except for short periods of time when rare ﬂoods generate very high boundary shear stresses. Furthermore, ephemerals experience an abundance of sediment supply while perennials usually suffer from limited sediment input. It is worth mentioning though that ephemeral streams from different regions encounter considerable variability in ﬂow/sediment conditions and exhibit a range of streambed material stratiﬁcations (e.g., Tooth, 2000; Hassan et al., 2006). The same is true for perennial streams (e.g., Whiting and King, 2003; Hassan et al., 2006). Grant suggests that the different timescales associated with the two stream types might be the driving mechanism resulting in the two distinct streambed structures. While it is difﬁcult to provide direct evidence in support or against this suggestion, indirect evidence exists indicating that ﬂood duration is important in enhancing the development of an armor layer but might not be the critical parameter identifying existence or absence of a coarser surface layer. For example, there are ephemeral streams that experience ﬂooding for sufﬁciently long time, up to a few days, and yet do not exhibit stratiﬁcation in terms of grain size (e.g., Greenbaum and Bergman, 2006). Furthermore, laboratory experiments, operating in sediment feed, sediment starved and recirculation modes, have shown that the bed material and bed load transport rates undergo signiﬁcant changes during the ﬁrst hour of the experiments (Profﬁtt, 1980; Diplas and Parker, 1985). However, in many cases, the starved sediment being the most noticeable ones, the degree of armoring becomes more pronounced with the duration of the experiment (Profﬁtt, 1980). One exception to the above statements is with regard to the duration of the receding portion of a ﬂood in ephemeral streams. A sufﬁciently long receding ﬂow could facilitate the establishment of an armor layer (Hassan et al., 2006). Strong evidence also exists about the importance of sediment supply on the make up of the surface layer (e.g., Whiting and King, 2003; Hassan et al., 2006). During the presentation, the authors suggested that the mechanism responsible for sediment transport during boundary shear stresses is much higher than those

312

P. Diplas, H. Shaheen

necessary for bed load transport, yet lower than those triggering sediment suspension, might resemble the ﬂuidization of the bed material. Contrary to the traditional approach where drag forces ﬂuidize the material, lift forces were proposed as responsible for ﬂuidizing the bed material in some ephemeral streams. In her discussion, Frostick, based on observations from experiments (Allan and Frostick, 1999), provides initial support for this concept. It is evident that carefully designed experiments will be needed to further examine this phenomenon. Hicks and Laronne generalize the concepts discussed in the streambed structure section by bringing to our attention observations from perennial braided rivers. They point out that some channels of perennial braided rivers exhibit ﬂow/sediment features similar to those observed in ephemeral streams while others those typically encountered in perennial streams. It is not therefore surprising that the former lack a well-established armor layer, which is present in the latter.

References Allan, A.F., Frostick, L., 1999. Framework dilation, winnowing, and matrix particle size: The behavior of some sand-gravel mixtures in a laboratory ﬂume. J. Sediment. Res. 69 (1), 21–26. Diplas, P., Parker, G., 1985. Pollution of gravel spawning grounds due to ﬁne sediment. Project Report No. 240, St. Anthony Falls Hydraulic Laboratory, University of Minnesota, Minneapolis, 192pp. Greenbaum, N., Bergman, N., 2006. Formation and evacuation of a large gravel-bar deposited during a major ﬂood in a Mediterranean ephemeral stream, Nahal Me’arot, NW Israel. Geomorphology 77, 169–186. Hassan, M.A., Egozi, R., Parker, G., 2006. Experiments on the effect of hydrograph characteristics on vertical grain sorting in gravel bed rivers. Water Resour. Res. 42 (9)10.1029/2005WR004707. Tooth, S., 2000. Process, form and change in dryland rivers: a review of recent research. Earth-Sci. Rev. 51, 67–107.

Gravel-Bed Rivers VI: From Process Understanding to River Restoration H. Habersack, H. Pie´gay, M. Rinaldi, Editors r 2008 Elsevier B.V. All rights reserved.

315

12 Non-stationarity of basin scale sediment delivery in response to climate change Tom J. Coulthard, John Lewin and Mark G. Macklin

Abstract Results from a cellular river basin evolution model (CAESAR) that contains a detailed multi-grain size sediment-transport model, indicate that over longer time scales (greater than 50 years) the relationship between sediment discharges (Qs) and water discharges (Qw) is not stable when routed through a prototype catchment. By modelling how the daily bedload yield from a medium sized river basin (383 km2) responds to changes in climate over the last 9000 years, high-resolution and long-term basin-scale bedload ratings curves can be simulated. These show that quasi-linear relationships can be established between Qs and Qw during stable climatic periods, but an increase in ﬂood magnitude and frequency over a sustained period (410 years) can lead to a very large increase in the amount of sediment delivered for identical sized ﬂoods. Thus different ‘climatic periods’ can produce signiﬁcantly different relationships between Qs and Qw. Furthermore, over the 9000 years simulated there is considerable variation in response, with daily sediment yields for a medium sized ﬂow (16–32 m3 s1) varying over eight orders of magnitude. The change in Qs/Qw relationship is believed to be conditioned by sediment supply and storage, and also by the ‘context’ of the climate period with respect to previous periods. These results have important implications for the application of sedimenttransport formulae, engineering calculations based on these (e.g., longer term reservoir siltation), and predictions of how river systems are likely to respond to climate change. 1.

Introduction

Understanding the relationship between catchment water and sediment discharges is important for comprehending how ﬂuvial systems operate. For geomorphologists controls on the delivery of sediment have numerous implications, from understanding ﬂuvial landform development to interpreting alluvial stratigraphies. And for E-mail address: [email protected] (T.J. Coulthard) ISSN: 0928-2025

DOI: 10.1016/S0928-2025(07)11131-7

316

T.J. Coulthard, J. Lewin, M.G. Macklin

engineers, knowledge of catchment sediment yields are vital, especially for assessing channel design, stability and sedimentation problems. Therefore, it is important for both scientists and practitioners to be able to model, and predict how the amounts of sediment transported from a catchment may respond to changes in water discharge over a range of time scales. To do this, researchers have measured catchment sediment yield – the amount of sediment that leaves a catchment over a measured period of time. This can be related to the size of the catchment and other factors such as climatic and altidudinal setting (e.g., Milliman and Syvitski, 1992). These records are variable, and changes in sediment yield over time have been related to changes in land use and land cover (Dearing, 1992) as well as to ﬂuctuations in climate (Wilby et al., 1997). More detailed studies have directly related catchment sediment yields (Qs) to catchment water discharge (Qw), and these are termed sediment ratings curves. The relationship between Qs and Qw is not straightforward and rarely linear. Typically, there is a hysteresis effect over a ﬂood, where there are reduced levels of sediment relative to water discharges in the latter parts of a ﬂood, due to sediment exhaustion. Sediment ratings curves are usually used to describe suspended sediment to Qw relationships, but bedload ratings curves have also been developed (Whiting et al., 1999; Emmett and Wolman, 2001; Ryan et al., 2002; Hayes et al., 2002; Whiting and King, 2003; Ryan et al., 2005). These also show considerable variation, but do not show the same daily hysteresis effects, though Moog and Whiting (1998) described a seasonal hysteresis. Both suspended sediment and bedload ratings curves are characterised by large quantities of scatter, which like hysteresis effects can be attributed to limitations in sediment supply. However, empirical models have been developed that use these ﬁeld data from ratings curves to predict both suspended sediment and bedload yields from catchments (Barry et al., 2004). A more sophisticated method is to use sediment-transport functions. These typically measure sediment discharge per unit width, and can therefore incorporate local channel characteristics such as slope, bed material etc. Therefore, they build in more of the local conditions and thus physics inﬂuencing sediment transport at that point. Sediment-transport functions have been extensively developed, but have encountered difﬁculties as they tend to perform well on the data with which they were developed, but less so when applied to different environments. Gomez and Church (1989) clearly illustrated this point with their comparison of 12 formulae, applied to eight different data sets. None performed accurately, the Bagnold equation was best, and only three others performed reasonably. The difﬁculties found in predicting bedload discharge are perhaps not surprising when the raw ﬁeld and laboratory data shows signiﬁcant variations (Reid et al., 1980; Ashmore, 1988; Hoey and Sutherland, 1991; Carling et al., 1998). For example, Cudden and Hoey (2003) measured bedload from two anabranches of a pro-glacial stream and noted substantial variations in bedload yields. These appear highly nonlinear, and Gomez and Phillips (1999) calculated that most of the non-linearity on 10,000 ﬂume bedload measurements could be described as chaotic. Causes for these irregularities include the threshold based nature of sediment entrainment and other threshold dominated processes operating on the bed of heterogeneous gravel-bed rivers, such as bed armouring and equal mobility. Furthermore, the passage of bedforms (e.g., dunes and bars) and larger scale features such as sediment waves

Non-stationarity of basin scale sediment delivery

317

(Nicholas et al., 1995) can cause irregularity in at-a-point bedload yields ranging from 1 to 1000s of m3 that may take seconds to years to pass. Over longer time scales, the impacts of climate or land cover change may also cause longer lasting perturbations in sediment delivery (Macklin and Lewin, 1989; Coulthard and Macklin, 2001). These problems are compounded by the physical difﬁculties associated in measuring bedload. Measurements are carried out using active methods (e.g., Helly Smith samplers, pressure pillows (e.g., Reid et al., 1980), magnetic devices (Tunnicliffe et al., 2000)) or passive techniques (sediment traps, repeat surveys of channel cut and ﬁll or lake/reservoir sedimentation rates). Active methods can give temporally highresolution data (hourly or less) but are demanding in human resources and typically only span a few hours or days. Passive techniques can provide much longer records, but with a far coarser temporal resolution, as data points are only collected when traps are emptied or cross sections re-surveyed. Therefore we are limited to either short high-resolution, or longer low-resolution records. Other researchers have looked to the geological record as a longer term recorder of changes in sediment yield. These demonstrate varying ﬂood-frequency episodes, with ‘wet’ and ‘dry’ periods related to climatic ﬂuctuations on a scale of tens to thousands of years (e.g., Macklin et al., 1992; Knox, 1993; Macklin and Lewin, 2003). Such timescales are compatible with the likely travel times of bed materials through catchment systems. However, the resolution of the dating methods used to establish ﬂuvial chronologies is generally not precise enough to accurately link sediment response to individual ﬂoods. Similarly, the resolution and accuracy of proxies used to reconstruct palaeo climate records also hampers correlation with a stratigraphy. Furthermore, the sedimentary record is palimpsest, a record that may have been re-worked and erased (Macklin and Lewin, 2003), and is not a straightforward ‘recorder’ being described as a ‘ﬁnicky’ by Paola (2003). Therefore, we are faced with the reality of having temporally limited data on present-day sediment yields (too short a length of record), and long-term records (too coarse a temporal resolution). As a consequence, it is difﬁcult to apply longer term records to shorter time scales (e.g., tens to hundreds of years) for instance to see how climate change may effect river systems. Similarly, it could be dangerous to apply present day process rates to interpret past longer term changes. Therefore, outside of our comparatively short length record we have little information on whether or how Qs/Qw relationships may change, and what might control it. Does the amount of sediment discharged for a given ﬂood change over time, if so by how much, and what may inﬂuence this? In order to better understand this and how river systems may respond to both individual ﬂoods, as well as longer term shifts in average ﬂood magnitude and frequency caused by climate changes, ideal data would be hourly water and sediment discharge data integrated across a river basin and spanning a period of several thousand years. This is clearly beyond the scope of monitoring studies and palaeoﬂood reconstruction techniques. But by using numerical modelling it is possible to explore the probable roles of changing ﬂood frequency and sediment supply and routing factors on a catchment wide basis. Indeed, previous studies have modelled non-linear sediment discharges from identical ﬂoods on small UK catchments

T.J. Coulthard, J. Lewin, M.G. Macklin

318

(Coulthard et al., 1998) and larger abstract landscapes over longer time scales (De Boer, 2001). They also show how decadal outputs of sediment can vary signiﬁcantly between catchments despite having identical drivers (Coulthard et al., 2005) as well as how thresholds are important for non-linear catchment response (Tucker, 2004). In this paper, we use a cellular landscape evolution model CAESAR (Coulthard et al., 2002) to reconstruct a series of bedload ratings curves based on 9000 year simulation of daily water and sediment discharges. The results presented here may have important implications for the management of catchments and some of the concepts that underpin our current understanding of geomorphology.

2.

Method

The results presented in the paper are based on further analysis of data ﬁrst presented in Coulthard et al. (2005), generated using the CAESAR landscape evolution model. A brief description of the model is provided below, but for a detailed account, readers are referred to Coulthard et al. (2002) and Van De Wiel et al. (2007). CAESAR is a cellular landscape evolution model (Coulthard, 2001; Willgoose, 2005) that represents a river catchment with a grid of equally sized square cells (as per a raster DEM). Each of these cells has properties, or values, for elevation, grain size, water discharge, ﬂow depth, vegetation cover etc. For every iteration or time step of the models operation these values are altered according to equations that describe four key groups of processes; hydrological, hydraulic, ﬂuvial erosion and deposition and slope processes. The hydrological model is an adaptation of TOPMODEL (Beven and Kirkby, 1979) and is driven by an hourly rainfall data set. Key parameters within TOPMODEL (m and K) can be altered to change the hydrograph peak and rate of ﬂood decay, and these are altered within CAESAR to simulate the effects of land cover change on catchment hydrology. This provides distributed values of soil-saturation levels, and when a grid cell is saturated, any excess discharge from the hydrological model is treated as surface ﬂow. This model is also dynamic, so wetted areas and thus the drainage network, can expand and contract during storm events. Surface discharge from the hydrological model is then fed into the hydraulic component, where depths, inundation areas and surface water routing are carried out. Flow depths are calculated for each cell using an adaptation of Mannings equation (using bed slope) and then routed according to a scanning multiple ﬂow routing algorithm. This sweeps across the catchment four times (north to south, east to west, west to east and south to north) routing water to the three cells in front, as per Murray and Paola (1994), using the slope between the water surface height and the receiving cells bed elevation. The maximum depth calculated for a cell through each of the four scans is recorded and taken as the cell ﬂow depth. Where a cell has a ﬂow depth, ﬂuvial erosion and deposition are carried out. Using water depth and bed slope between cells, bedload transport for nine separate grain sizes (1–256 mm in Phi classes) is calculated using the Einstein–Brown (Einstein, 1950) formulae. Erosion and deposition of the separate size fractions is implemented through the use of an active layer system (Hoey and Ferguson, 1994) where a

Non-stationarity of basin scale sediment delivery

319

proportion of each fraction is held in a series of layers that during erosion can be replenished from layers below, or during deposition displace material to lower layers. Importantly, this system allows many of the processes associated with heterogeneous bed sediment transport to be modelled, such as bed armouring and size supply limited entrainment. Two sets of slope processes are modelled. First, mass movement occurs when a threshold slope angle is exceeded and secondly, soil creep is calculated according to local slopes. Importantly, both slope processes are fully integrated within the ﬂuvial model, which allows the addition of sediment into the ﬂuvial system from a landslide or eroding river bank, for example. In summary, CAESAR allows hydrological, hydraulic, ﬂuvial and slope processes to be simulated over an entire catchment modelling series of individual ﬂood events, driven by an hourly rainfall record and DEM. The simulations described here were carried out on the upper part of the River Swale, UK, the northern tributary of the Yorkshire Ouse system. The modelled catchment area covers 383 km2 and varies in relief from 514 to 20 m A.O.D. The Swale has a mixed lithology of Carboniferous sandstone, gritstone and limestone, and was glaciated during the Quaternary, leaving wide steep walled valleys covered with periglacial and glacial deposits. The catchment was extensively forested (Flemming, 1998) but present day land cover is typically rough grazing on upland moorland, with pasture on the lower sections and valley ﬂoors. The catchment is sufﬁciently high to be unaffected by Holocene sea-level rise. The model was applied to a 50 m resolution DEM of the Swale, and was driven by a proxy climate and land cover data set covering the last 9000 years. The topography of the Swale catchment 9000 years ago is impossible to reconstruct, so rather than make crude reconstructions based on limited data, the present day topography was used. This is not a wholly unreasonable assumption, as ﬁeld evidence indicates that there has been less than 2–4 m of vertical channel and valley ﬂoor movement over the Holocene. For such a long simulation, a 9000 year hourly rainfall record was required. We used a combination of two climatic indices derived from peat bog surface wetness reconstructions from northern England for the period 6300 cal. BP to present (from Barber et al., 1994) and from 9000 to 6300 from Scotland (Anderson et al., 1998). A full description and discussion of this method can be found in Coulthard and Macklin (2001) and Coulthard et al. (2005). These indices provided climate data in 50 year steps and were then normalised between 0.75 and 2.25 to create a rainfall index. The model was then driven by a 10 year hourly rainfall record that is repeated ﬁve times to span 50 years, then multiplied by the rainfall index to create 9000 year proxy hourly rainfall data. For land-cover changes, a basic reconstruction of catchment deforestation constructed from local palynological records were used to change the ‘m’ parameter in the hydrological model.

3.

Results

The simulation generated a time series of 87 106 data points of hourly ﬂow and sediment discharge. To simplify analysis of the data set, it was divided into 50 year

T.J. Coulthard, J. Lewin, M.G. Macklin

320

sections, discharge data were averaged and bedload yields summed to create a daily time series. In effect, we have used the model to generate daily water discharge and bedload totals for the River Swale over a 9000 year period. Fig. 12.1 shows the relationship between Qw (water discharge) and Qs (sediment yield) for three 50 year sections (50–100, 100–150 and 150–200 cal. BP). Despite being plotted on log–log axis these data demonstrate a signiﬁcant amount of scatter, with an interesting increase in variability to up to eight orders of magnitude around the 15–50 m3 s1 Qw. These periods were chosen as they contained very high (150–200 cal. BP), high (100–150 cal. BP) and medium (50–100 cal. BP) sediment discharges, though similar patterns of scatter are found in all 50 year sections. When the data is grouped

1000000

1000000

1000000

100000

100000

100000

10000

10000

10000

1000

1000

1000

100

100

100

10

10

10 1

1 0.01

0.1

0.1

1

10

100

1000 0.01

0.01

0.1

1 1 0.1

10

100 1000

0.1

0.1

0.01

0.01

0.001

0.001

0.0001

0.0001

0.001 0.0001

0.00001

0.00001 0.000001

0.000001

0.000001

1

100

10

1000

0.00001

1000000 y = 37.833x + 323.58 R2 = 0.0422

150

100000 10000

100 50

Daily sediment yield (m3)

Linear (50)

1000

Linear (150)

100

Linear (100) y = 6.4567x - 16.271 R2 = 0.5217

10 1 0.1

y = 2.9363x - 17.33 R2 = 0.306

0.01 0.001 0.0001 0.00001

0.000001 0.01

0.1

1

10

100

1000

Mean Daily Flow (m3 sec -1) Figure 12.1. This ﬁgure illustrates Qw and Qs plotted for three 50-year sections of the data set, 50–100, 100–150 and 150–200 cal. BP. Note, both axis are log scaled. The top three plots show each data set (50, 100 and 150) plotted separately.

Non-stationarity of basin scale sediment delivery

321

into log classes according to Qw (1, 2, 4, 8, 16, 32, 64, 128, 256 and 512 respectively) the standard deviation of the Qs in each bin varied considerably, with a peak in the 16–32 m3 s1 group. Using this group from the 150–200 cal. BP section and classing the Qs into linear bins (incrementing from 0 to 20,000 in steps of 1000) the frequency of daily sediment yield (Qs) to water discharges in the 16–32 m3 s1 group can be plotted as a power law (Fig. 12.2b). Data from 50 cal. BP period, with far lower sediment yields, also follow a similar relationship, though smaller bin sizes are required (Fig. 12.2a). Examining Fig. 12.1, there is also considerable difference in the angle of the regression lines plotted through the data sets. This suggests that there could be signiﬁcant variation in the Qs/Qw relationship during different periods of the model operation. However, R2 values indicate that the relationship is clearly not signiﬁcant due to the large amount of scatter. In order to reduce this scatter and to allow comparisons in the Qs/Qw relationship between each 50 year section of the simulation, these data were grouped into log classes as above. When the medians of Qs for each of these bins were plotted these revealed a good linear relationship as shown in Fig. 12.3. This good level of ﬁt was found to be consistent for each of the 50 year sections of data, though it must be remembered that using medians conceals a large amount of scatter and therefore should only be used as a comparator between individual 50 year sections.

10000 150 50 Power (150)

1000

Frequency

Power (50)

100 y = 235347x -1.9608 R2 = 0.9345

10

50-100 cal. BP

y = 5E+08x -1.9822 R 2 = 0.9154

150-200 cal. BP

1 10

100

1000 Daily Bedload Yield

10000

100000

(m3)

Figure 12.2. Power law distribution of bedload yield for (a) time period 150–200 cal. BP and (b) 50–100 cal. BP.

T.J. Coulthard, J. Lewin, M.G. Macklin

322 1400 1800

Median daily sediment yield (m3)

1200

y = 8.7701x - 71.444 R2 = 0.9406

1850 1900

1000

Linear (1900) Linear (1850)

800

Linear (1800)

y = 4.7874x - 45.638 R2 = 0.9319

600 400 y = 1.2572x - 2.9453 R2 = 0.9709

200 0 0

20

40

60

80

100

120

140

Mean Daily Flow (m3 s -1) Figure 12.3. Medians of simulated sediment yields for the periods 1800, 1850 and 1900 cal. BP.

Linear regression was then carried out for the medians of all 50 year time sections and the slopes of the regression line are shown as a time series in Fig. 12.4. This indicates how the angle of the modelled relationship between Qs/Qw changes quite dramatically throughout the simulation as well as following sediment yield and climate input closely. The standard deviation and median for the 16–32 m3 s1 class were also calculated and are shown in temporal sequence for all data in Fig. 12.5.

4. 4.1.

Discussion Scatter

Fig. 12.1 indicates that there is considerable scatter within the Qs/Qw relationship, and it is apparent that there is a ﬂow size with substantially more scatter than others (16–32 m3 s1). This is not the most frequent class of ﬂow (8–16 m3 s1) and this area of scatter seems to persist throughout every 50 year section, though for some sections is less well deﬁned. These results raise two questions: why is there so much scatter in the Qs and why does there appear to be a ﬂow range that produces substantially more scatter than any other? At ﬁrst the large amounts of scatter may seem to be inappropriate, as CAESAR is driven by a conventional sediment-transport formulation (Einstein, 1950) that has no stochastic element. Thus it might be expected that sediment outputs would follow this relationship. However, within CAESAR this formulae is applied to individual cells, with sediment transport calculated locally, based upon the ﬂow depths and bed

Non-stationarity of basin scale sediment delivery

323 100

2.5

10

Climate proxy

2

1

1.5

0.1

1

0.5

0.01

0 0

1000

2000

3000

4000

5000

6000

7000

8000

Slope of Qs Q w relationship

Climate Slope of Qs Qw

0.001 9000

Years Cal. BP 100000000

100 Sediment yield Slope of Qs Qw

10000000

100000

1

10000 0.1

1000 100

Slope of Qs Qw relationship

Sediment Yield (m3 per 50 yrs)

10 1000000

0.01 10 1 0

1000

2000

3000

4000

5000

6000

7000

8000

0.001 9000

Years Cal. BP Figure 12.4. Slope of Qs/Qw relationship plotted with the climate driver for the simulations (top) and the modelled sediment yield (bottom).

slopes of each cell. The output presented here, is the product of erosion and deposition being carried out over thousands of grid cells within the model. Therefore, by calculating sediment transport on a cell by cell basis, CAESAR mimics the processes of internal storage and re-mobilisation of sediment that can operate within real catchments. Furthermore, as CAESAR contains a sophisticated sediment-transport model using several grain sizes, when supply of a certain grain size is exhausted within a cell, no more of this fraction can be eroded. Thus sediment supply

T.J. Coulthard, J. Lewin, M.G. Macklin

324 100000000

10000 Sediment yield Standard deviation

1000

Median of 16-32 flow class

1000000

100 100000 10 10000 1 1000 0.1 100

Standard Deviation and Median

Sediment Yield (m3 per 50 years)

10000000

0.01

10 1 0

1000

2000

3000

4000

5000

6000

7000

8000

0.001 9000

Years Cal. BP

Figure 12.5. Standard deviation and medians of the 16–32 m3 s1 ﬂood class plotted with the modelled sediment yield.

limitations, such as bed armouring and selective entrainment are incorporated in each cell. This can generate internal erosional thresholds, so sediment is not eroded until stream powers are great enough to move the particle itself or to remove the coarser armour layer above. Therefore, erosion is threshold dominated giving rise to high variations in sediment delivery. Furthermore, as the model incorporates slope processes (soil creep and mass movement) material can be added from landslides, which are also threshold controlled, again introducing high variations in sediment delivery. Explanations for why one class of mean daily ﬂows produces such high levels of variability are not straightforward. It is logical to assume that in threshold based supply limited catchments, small ﬂoods will largely generate little sediment yield and large ﬂoods will generally mobilise a substantial volume of material. Therefore the medium sized ﬂows, that here show most variability, may be close to erosional thresholds, where there is high potential for variability. In order to evaluate what might control the variability shown in the model results, data from similar simulations on two nearby catchments were examined. Fig. 12.6 plots Qs/Qw relationships from the River Swale, as well as the nearby River Nidd (281 km2) and River Wharfe (697 km2). Both exhibit similar patterns of scatter, with the similar sized Nidd being closest. However, the larger Wharfe catchment shows greatest variability in a larger ﬂood size class. This suggests that the size of the catchment or possibly its morphology may be a controlling factor. Possibly it is not small scale erosional thresholds such as bed armour breaching that inﬂuence the scatter in this way, but larger controls on thresholds, such as valley ﬂoor shape and internal catchment storage. Further research is required into the precise causes of this behaviour, and it is possible that other factors, for example, grain size and land cover may strongly inﬂuence this.

Non-stationarity of basin scale sediment delivery

325

100000 10000

nidd wharfe

1000

swale

Daily Sediment Yield (m 3)

100 10 1 0.1 0.01 0.001 0.0001 0.00001

0.000001 0.01

0.1

1

10

100

1000

Mean Daily Flow (m3 s -1) Figure 12.6. Qs/Qw plots for the rivers Swale, Wharfe and Nidd.

Fig. 12.2b shows that the distribution of the scatter follows a power law. This indicates that Qs may vary from 0 to 20,000 m3day1 for a mean daily ﬂow between 16 and 32 m3 s1. Interestingly, Fig. 12.2a indicates that different climatic periods produce different power law relationships, showing that whilst the magnitude of scatter may be different, they both follow similar distributions. Power laws can also be indicative of systems exhibiting self organised criticality (SOC) though this is not an area we wish to explore in this paper. 4.1.1.

Implications of scatter

The large amounts of scatter (up to eight orders of magnitude for ﬂows between 16 and 32 m3 s1) suggest that bedload ratings curves are not ideal for predicting sediment discharges. Sediment-transport functions that take account of factors such as channel width, depth and local slopes where the bedload is measured (here the edge of the DEM) could improve these predictions. But, both will still feel the effects of changes in sediment supply outlined above and this study highlights some of the difﬁculties faced when trying to predict sediment transport. An alternative method is to smooth these data, and average it over monthly, annual or even decadal time steps. By using longer measurement intervals we might then be able to derive generalised relationships between long-term ﬂow and sediment yield data. But such relationships, though statistically less variable would mask the extremes of response that might be crucial for management strategies. The clustering of most scatter around a certain ﬂow size may be of great importance to engineers and practitioners. Akin to the concept of the effective ﬂood (Wolman and

T.J. Coulthard, J. Lewin, M.G. Macklin

326

Miller, 1960) is there a size of ﬂood that may have extra levels of uncertainty associated with it? Certainly, these results suggest that this may be the case, and that such magnitude of ﬂoods are very common. However, when reviewing these data presented here, it should be remembered that Figs. 12.1 and 12.6 do not represent individual ﬂood totals, they are daily sediment totals compared to daily discharge averages. Therefore, these data could easily conceal daily/event scale hysteresis effects. Furthermore, the data for 150–200 years cal. BP represents an extreme of wetness and the greatest level of variation. Under dryer conditions the maximum size of variation is far less (e.g., see Fig. 12.2b) but the amount of scatter remains very high.

4.2.

Non-stationarity of response

Figs. 12.4 and 12.5 show how 50 year climatic ‘periods’ produce a different relationship between Qs and Qw. There is an increase in the gradient of the relationship with increasing wetness, resulting in a greater Qs for the same Qw in a wetter climate (represented by an increase in the climate index). In a wetter period in a mid latitude temperate river basin, it would be anticipated that there would be a higher overall sediment yield, due to larger ﬂoods. But the slope shows how the ratings curve, the Qs/Qw relationship, changes signiﬁcantly. In effect the relationship is non-stationary over time. These changes can be explained ﬁrstly by morphological changes within the catchment. Increasing ﬂood size leads to an expansion of the drainage network, that generates fresh material from expanding stream heads and incision within ﬁrst-order streams as demonstrated by previous applications of CAESAR (Coulthard et al., 2002). In a small catchment, it might be expected that all this new material would be ﬂushed from the system. But within a medium to large sized catchment, such as the Swale, this material will be deposited and progressively moved downstream as a series of sediment waves or slugs. This in turn leads to the increase in the Qs/Qw relationship, as the larger ﬂoods have mobilised sediment that is then available for smaller ﬂoods to export. In effect, larger ﬂoods or wetter climatic periods loosen or release large amounts of sediment, increasing the volume of material that could be transported for smaller ﬂood events. Similarly, changes in the general pattern of ﬂood events may breach armour layers on the stream bed, which can release sediment as well as make the channel more vulnerable to incision (Coulthard and Van De Wiel, in press). The sudden shift in Qs/Qw could also be explained by changes in the channel pattern. Signiﬁcant increases in bedload yield would be consistent with a transformation from a single thread channel to a braided one. This can be seen in Fig. 12.7, which shows the channel pattern for two identical size ﬂoods at 4500 and 200 years cal. BP, corresponding with low and high Qs/Qw relationships. The upper 4500 cal. BP image shows a predominantly single channel, with some overbank inundation cutting across meander bends. The lower 200 cal. BP image shows a widening of the channel in the upper parts (as an adjustment to increased ﬂows) and a shift in channel pattern to a multi-threaded braided planform. Such behaviour is consistent with ﬁeld evidence from similar catchments in the UK (the river South Tyne;

Non-stationarity of basin scale sediment delivery

327

Figure 12.7. Areas inundated by a 50 m3 s1 ﬂow at 4500 (top) and 200 (bottom) cal. BP.

Macklin and Lewin, 1989) that showed dramatic changes in channel pattern and valley ﬂoor aggradation. Such a planform change could also explain increases in scatter and the standard deviation (Fig. 12.6) during wetter climates, as braided channels can have a far greater ﬂuctuating bedload yield.

T.J. Coulthard, J. Lewin, M.G. Macklin

328

The changes in the Qs/Qw relationship can also be explained by the effects of sediment recharge and removal. For example, during several hundred years of relatively constant moderate rainfall events, the river channel armours and adjusts its width/depth ratio’s to accommodate the ﬂow. During this period, in upland areas soil creep will slowly move soil and sediment to the base of slopes and thus the margins of streams. Present sediment supply remains the same, but potential supply has increased. If there is then a wetter climatic period, with more frequent and larger ﬂoods, the relative equilibrium within the channel will be broken and erosion will start to remove this available material adjacent to channels. This increases the relative sediment delivery per ﬂood size, as we see in Fig. 12.1.

4.2.1.

Implications

These results have signiﬁcant implications for the application of bedload rating curves as they show that the Qs/Qw relationship can vary considerably over time. The modelling results presented here show that there is a double response, as not only will wetter climate periods increase sediment yields through larger ﬂoods, but will also increase the relative sediment yield for all ﬂoods (the Qs/Qw relationship). This is especially important for design predictions as if, for example, we predict the rates of sedimentation in a reservoir – or around a structure such as a bridge, based on existing sediment/bedload ratings curves, these may alter signiﬁcantly in the future. Therefore future changes in climate may inﬂuence not only the size of future ﬂoods, but also the relative volumes of sediment released. This is especially pertinent given the rapid present day changes to our climate. For sedimentologists, when interpreting sequences, the double effect may lead to a large deposit being disproportionately thick, as there is clearly a far from linear relationship between climate and unit thickness. Relating both scatter and the Qs/Qw relationship, Fig. 12.6 shows how an increase in the standard deviation of the 16–32 m3 s1 ﬂood class follows the climate and sediment yield. This indicates that not only are Qs/Qw values going to change, but also the variability, which could make engineering decisions and sedimentological interpretations even harder. Finally, as changes in the Qs/Qw relationship are inﬂuenced by the release of stored sediment and the renewal of erosional activity in stream heads and adjacent parts of the channel, then catchment response will be heavily inﬂuenced by the previous patterns of erosion and deposition. Therefore, catchment sediment yields are heavily contingent on the system history and predictions, and assumptions based upon bedload ratings curves should take this into account. Preliminary analysis examining how the Qs/Qw relationship varies during the 50 year blocks studied here indicates that there is signiﬁcant adjustment over the ﬁrst 20 years followed by more minor changes. This suggests that the reaction time to climatic changes of the simulated River Swale is fairly rapid, but how this timing scales between catchments of different shapes and morphologies is beyond the scope of this paper and warrants further study. However, this does give us some indication of how rapidly river systems may respond to climatic changes.

Non-stationarity of basin scale sediment delivery 5.

329

Conclusions

Catchment modelling of bedload sediment yield for a moderate sized river basin over the last 9000 years using the CAESAR model suggests that the spatial and temporal complexities of sediment transport and storage produce markedly non-linear relationships between water discharge and sediment yield. The yield of any particular discharge is conditioned by prior events, sediment availability and sediment supply. This is particularly striking for moderate sized ﬂows (in the case of the modelled River Swale 16–32 m3 s1) for which daily yields vary over eight orders of magnitude. For a given ﬂow magnitude, wetter periods also yield more sediment than drier ones, probably because sediment mobilisation then exceeds critical thresholds, as in the breaching of bed-armouring, landslide generation or storage activation. The spatial complexities of larger catchments, in terms of available storage and the temporal routing of sediment transfers would also appear to make predictions of Qs/Qw relationships much more variable than in the case of small catchments. Transitions between ‘wetter’ and ‘drier’ periods also seem liable to particularly uncertain predictions. All these uncertainties have serious management implications at times when rapid climate change and human impacts on sediment yields are much in evidence. There are no long-term data sets available with which to attempt model validation, though there are no reasons to suppose that the use of a conventional sediment-transport equation, along with their hydrological and process relationships, to model catchment wide transportation and delivery, is in any general sense invalid. Indeed, it suggests that ‘site’ transport equations do need to be aggregated with procedures which incorporate catchment behaviour, if sediment yield variability is to be simulated in an apparently realistic manner. Modelling reveals some interesting characteristics of this variability which have management consequence at times of rapid environmental change.

Acknowledgements We thank Chris Paola, Jim Pizzuto and a third anonymous referee for their excellent comments and guidance in the preparation of this paper. T.J. Coulthard also thanks the organisers of GBR6 for the opportunity to present this research.

References Anderson, D.E., Binney, H.A., Smith, M.A., 1998. Evidence for abrupt climatic change in northern Scotland between 3900 and 3500 calendar years BP. Holocene 8, 97–103. Ashmore, P.E., 1988. Channel morphology and bedload pulses in braided, gravel-bed streams. Geograﬁska Annaler 73 (A1), 37–52. Barber, K.E., Chambers, F.M., Maddy, D., et al., 1994. A sensitive high resolution record of late Holocene climatic change from a raised bog in northern England. Holocene 4, 198–205.

330

T.J. Coulthard, J. Lewin, M.G. Macklin

Barry, J.J., Bufﬁngton, J.M., King, J.G., 2004. A general power equation for predicting bed load transport rates in gravel bed rivers. Water Resour. Res. doi:10.1029/2004WR003190. Beven, K.J., Kirkby, M.J., 1979. A physically based variable contributing-area model of catchment hydrology. Hydrol. Sci. Bull. 24, 43–69. Carling, P.A., Williams, J.J., Kelsey, A., et al., 1998. Coarse bedload transport in a mountain river. Earth Surf. Process. Landf. 23, 141–157. Coulthard, T.J., 2001. Landscape evolution models: A software review. Hydrol. Process. 15, 165–173. Coulthard, T.J., Van De Wiel, M.J., in press. Quantifying ﬂuvial non-linearity and ﬁnding self organized criticality? Insights from simulations of river basin evolution. Geomorphology. Coulthard, T.J., Kirkby, M.J., Macklin, M.G., 1998. Non-linearity and spatial resolution in a cellular automaton model of a small upland basin. Hydrol. Earth Syst. Sci. 2, 257–264. Coulthard, T.J., Lewin, J., Macklin, M.G., 2005. Modelling differential catchment response to environmental change. Geomorphology 69, 222–241. Coulthard, T.J., Macklin, M.G., 2001. How sensitive are river systems to climate and land-use changes? A model-based evaluation. J. Quaternary Sci. 16, 347–351. Coulthard, T.J., Macklin, M.G., Kirkby, M.J., 2002. A cellular model of Holocene upland river basin and alluvial fan evolution. Earth Surf. Process. Landf. 27, 269–288. Cudden, J.R., Hoey, T.B., 2003. The causes of bedload pulses in a gravel channel: The implications of bedload grain-size distributions. Earth Surf. Process. Landf. 28, 1411–1428. De Boer, D.H., 2001. Self-organisation in ﬂuvial landscapes: Sediment dynamics as an emergent property. Computers Geosci. 27, 995–1003. Dearing, J.A., 1992. Sediment yields and sources in a Welsh upland lake-catchment during the last 800 years. Earth Surf. Process. Landf. 17, 1–22. Einstein, H.A., 1950. The bed-load function for sediment transport on open channel ﬂows. Tech. Bull. No. 1026, USDA, Soil Conservation Service, 71. Emmett, W.M., Wolman, M.G., 2001. Effective discharge and gavel bed rivers. Earth Surf. Process. Landf. 26, 1369–1380. Flemming, A., 1998. Swaledale: Valley of the Wild River, Edinburgh University Press, pp. 138–140. Gomez, B., Church, M., 1989. An assessment of bed load sediment transport formulae for gravel bed rsivers. Water Res. Res. 25 (6), 1161–1186. Gomez, B., Phillips, J.D., 1999. Deterministic uncertainty in bed load transport. J. Hydraul. Eng. 125, 305–308. Hayes, S.K., Montgomery, D.R., Newhall, C.G., 2002. Fluvial sediment transport and deposition following the 1991 eruption of Mount Pinatubo. Geomorphology 45, 211–224. Hoey, T.B., Ferguson, R., 1994. Numerical simulation of downstream ﬁning by selective transport in gravel bed rivers: Model development and illustration. Water Resour. Res. 30, 2251–2260. Hoey, T.B., Sutherland, A.J., 1991. Channel morphology and bedload pulses in braided rivers: A laboratory study. Earth Surf. Process. Landf. 16, 447–462. Knox, J.C., 1993. Large increases in ﬂood magnitude in response to modest changes in climate. Nature 361, 430–432. Macklin, M.G., Lewin, J., 1989. Sediment transfer and transformation of an alluvial valley ﬂoor: The river south Tyne, Northumbria, UK. Earth Surf. Process. Landf. 14, 233–246. Macklin, M.G., Lewin, J., 2003. River sediments, great ﬂoods and centennial scale Holocene climate change. J. Quat. Sci. 18, 101–105. Macklin, M.G., Rumsby, B.T., Heap, T., 1992. Flood alluviation and entrenchment: Holocene valley ﬂoor development and transformation in the British uplands. Geol. Soc. Am. Bull. 104, 631–643. Milliman, J.D., Syvitski, J.P.M., 1992. Geomorphic/tectonic control of sediment discharge to the ocean: The importance of small mountainous rivers. J. Geol. 100, 524–525. Moog, D.B., Whiting, P.J., 1998. Annual hysteresis in bed load rating curves. Water Resour. Res. 34 (9), 2393–2399. Murray, A.B., Paola, C., 1994. A cellular model of braided rivers. Nature 371, 54–57. Nicholas, A.P., Ashworth, P.J., Kirkby, M.J., et al., 1995. Sediment slugs: Large scale ﬂuctuations in ﬂuvial sediment transport rates and storage volumes. Prog. Phys. Geogr. 19 (4), 500–519.

Non-stationarity of basin scale sediment delivery

331

Paola, C., 2003. Floods of record. Nature 425, 459. Reid, I., Layman, J.T., Frostick, L.E., 1980. The continuous measurement of bedload discharge. J. Hydraul. Res. 18 (3), 243–249. Ryan, S.E., Porth, L.S., Troendle, C.A., 2002. Deﬁning phases of bedload transport using piecewise regression. Earth Surf. Process. Landf. 27, 971–990. Ryan, S.E., Porth, L.S., Troendle, C.A., 2005. Coarse sediment transport in mountain streams in Colorado and Wyoming, USA. Earth Surf. Process. Landf. 30, 269–288. Tucker, G.E., 2004. Drainage basin sensitivity to tectonic and climatic forcing: Implications of a stochastic model for the role of entrainment and erosion thresholds. Earth Surf. Process. Landf. 29, 185–205. Tunnicliffe, J., Gottesfeld, A.S., Mohamed, M., 2000. High resolution measurement of bedload transport. Hydrol. Process. 14, 2631–2643. Van De Wiel, M.J., Coulthard, T.J., Macklin, M.G., Lewin, J., 2007. Embedding reach-scale ﬂuvial dynamics within the CAESAR cellular automaton landscape evolution model. Geomorphology. doi:10.1016/j.geomorph.2006.10.024. Whiting, P.J., King, J.G., 2003. Surface particle sizes on armoured gravel streambeds: Effects of supply and hydraulics. Earth Surf. Process. Landf. 28, 1459–1471. Whiting, P.J., Stamm, J.F., Moog, D.B., Orndorff, R.L., 1999. Sediment transporting ﬂows in headwater streams. GSA Bull. 111 (3), 450–466. Wilby, R.L., Dalgleish, H.L., Foster, I.D.L., 1997. The impact of weather patterns on historic and contemporary catchment sediment yields. Earth Surf. Process. Landf. 22, 353–363. Willgoose, G., 2005. Mathematical modelling of whole landscape evolution. Ann. Review Earth Planetary Sci. 33, 433–459. Wolman, M.G., Miller, J.P., 1960. Magnitude and frequency of forces in geomorphic processes. J. Geol. 68, 54–74.

Discussion by Murray Hicks I note the general similarity between widely varying behaviour of the supply- and storage-conditioned bedload transport ratings that appeared from the CAESAR modelling and the rating relations published recently by Barry et al. (2004). The latter are based on empirical analysis of river datasets from a variety of basins, and the coefﬁcients of the ratings appear to vary according to basin scale and basin characteristics that condition bed-material supply. In essence, the Barry et al. results may provide some veriﬁcation of the CAESAR results if they are recast in the form of Barry et al.’s relation. My suggestion is to make this comparison.

Reference Barry, J.J., Bufﬁngton, J.M., King, J.G., 2004. A general power equation for predicting bed load transport rates in gravel bed rivers. Water Resour. Res. 40, W10401.

Reply by the authors Firstly, I would like to thank Murray for drawing our attention to the work of Barry et al. (2004) which indeed provides a useful comparator for these results. In their paper, Barry et al. develop a simple sediment-transport function based on a power law relationship between total sediment transport per unit width and discharge.

Non-stationarity of basin scale sediment delivery

331

Paola, C., 2003. Floods of record. Nature 425, 459. Reid, I., Layman, J.T., Frostick, L.E., 1980. The continuous measurement of bedload discharge. J. Hydraul. Res. 18 (3), 243–249. Ryan, S.E., Porth, L.S., Troendle, C.A., 2002. Deﬁning phases of bedload transport using piecewise regression. Earth Surf. Process. Landf. 27, 971–990. Ryan, S.E., Porth, L.S., Troendle, C.A., 2005. Coarse sediment transport in mountain streams in Colorado and Wyoming, USA. Earth Surf. Process. Landf. 30, 269–288. Tucker, G.E., 2004. Drainage basin sensitivity to tectonic and climatic forcing: Implications of a stochastic model for the role of entrainment and erosion thresholds. Earth Surf. Process. Landf. 29, 185–205. Tunnicliffe, J., Gottesfeld, A.S., Mohamed, M., 2000. High resolution measurement of bedload transport. Hydrol. Process. 14, 2631–2643. Van De Wiel, M.J., Coulthard, T.J., Macklin, M.G., Lewin, J., 2007. Embedding reach-scale ﬂuvial dynamics within the CAESAR cellular automaton landscape evolution model. Geomorphology. doi:10.1016/j.geomorph.2006.10.024. Whiting, P.J., King, J.G., 2003. Surface particle sizes on armoured gravel streambeds: Effects of supply and hydraulics. Earth Surf. Process. Landf. 28, 1459–1471. Whiting, P.J., Stamm, J.F., Moog, D.B., Orndorff, R.L., 1999. Sediment transporting ﬂows in headwater streams. GSA Bull. 111 (3), 450–466. Wilby, R.L., Dalgleish, H.L., Foster, I.D.L., 1997. The impact of weather patterns on historic and contemporary catchment sediment yields. Earth Surf. Process. Landf. 22, 353–363. Willgoose, G., 2005. Mathematical modelling of whole landscape evolution. Ann. Review Earth Planetary Sci. 33, 433–459. Wolman, M.G., Miller, J.P., 1960. Magnitude and frequency of forces in geomorphic processes. J. Geol. 68, 54–74.

Discussion by Murray Hicks I note the general similarity between widely varying behaviour of the supply- and storage-conditioned bedload transport ratings that appeared from the CAESAR modelling and the rating relations published recently by Barry et al. (2004). The latter are based on empirical analysis of river datasets from a variety of basins, and the coefﬁcients of the ratings appear to vary according to basin scale and basin characteristics that condition bed-material supply. In essence, the Barry et al. results may provide some veriﬁcation of the CAESAR results if they are recast in the form of Barry et al.’s relation. My suggestion is to make this comparison.

Reference Barry, J.J., Bufﬁngton, J.M., King, J.G., 2004. A general power equation for predicting bed load transport rates in gravel bed rivers. Water Resour. Res. 40, W10401.

Reply by the authors Firstly, I would like to thank Murray for drawing our attention to the work of Barry et al. (2004) which indeed provides a useful comparator for these results. In their paper, Barry et al. develop a simple sediment-transport function based on a power law relationship between total sediment transport per unit width and discharge.

Non-stationarity of basin scale sediment delivery

331

Paola, C., 2003. Floods of record. Nature 425, 459. Reid, I., Layman, J.T., Frostick, L.E., 1980. The continuous measurement of bedload discharge. J. Hydraul. Res. 18 (3), 243–249. Ryan, S.E., Porth, L.S., Troendle, C.A., 2002. Deﬁning phases of bedload transport using piecewise regression. Earth Surf. Process. Landf. 27, 971–990. Ryan, S.E., Porth, L.S., Troendle, C.A., 2005. Coarse sediment transport in mountain streams in Colorado and Wyoming, USA. Earth Surf. Process. Landf. 30, 269–288. Tucker, G.E., 2004. Drainage basin sensitivity to tectonic and climatic forcing: Implications of a stochastic model for the role of entrainment and erosion thresholds. Earth Surf. Process. Landf. 29, 185–205. Tunnicliffe, J., Gottesfeld, A.S., Mohamed, M., 2000. High resolution measurement of bedload transport. Hydrol. Process. 14, 2631–2643. Van De Wiel, M.J., Coulthard, T.J., Macklin, M.G., Lewin, J., 2007. Embedding reach-scale ﬂuvial dynamics within the CAESAR cellular automaton landscape evolution model. Geomorphology. doi:10.1016/j.geomorph.2006.10.024. Whiting, P.J., King, J.G., 2003. Surface particle sizes on armoured gravel streambeds: Effects of supply and hydraulics. Earth Surf. Process. Landf. 28, 1459–1471. Whiting, P.J., Stamm, J.F., Moog, D.B., Orndorff, R.L., 1999. Sediment transporting ﬂows in headwater streams. GSA Bull. 111 (3), 450–466. Wilby, R.L., Dalgleish, H.L., Foster, I.D.L., 1997. The impact of weather patterns on historic and contemporary catchment sediment yields. Earth Surf. Process. Landf. 22, 353–363. Willgoose, G., 2005. Mathematical modelling of whole landscape evolution. Ann. Review Earth Planetary Sci. 33, 433–459. Wolman, M.G., Miller, J.P., 1960. Magnitude and frequency of forces in geomorphic processes. J. Geol. 68, 54–74.

Discussion by Murray Hicks I note the general similarity between widely varying behaviour of the supply- and storage-conditioned bedload transport ratings that appeared from the CAESAR modelling and the rating relations published recently by Barry et al. (2004). The latter are based on empirical analysis of river datasets from a variety of basins, and the coefﬁcients of the ratings appear to vary according to basin scale and basin characteristics that condition bed-material supply. In essence, the Barry et al. results may provide some veriﬁcation of the CAESAR results if they are recast in the form of Barry et al.’s relation. My suggestion is to make this comparison.

Reference Barry, J.J., Bufﬁngton, J.M., King, J.G., 2004. A general power equation for predicting bed load transport rates in gravel bed rivers. Water Resour. Res. 40, W10401.

Reply by the authors Firstly, I would like to thank Murray for drawing our attention to the work of Barry et al. (2004) which indeed provides a useful comparator for these results. In their paper, Barry et al. develop a simple sediment-transport function based on a power law relationship between total sediment transport per unit width and discharge.

T.J. Coulthard, J. Lewin, M.G. Macklin

332 These take the form qb ¼ cQe

Where qb is the transport per unit width (in kg s1 m1), c is a coefﬁcient, Q discharge and e an exponent. They calculated the coefﬁcient and exponent based on 2104 bedload measurements in 24 different gravel-bed rivers in Idaho, and found that the coefﬁcient varied from 110 to 12, and the exponent from 1.5 to 3.8. Furthermore the exponent could be related to another factor (q*) describing the level of bed armour and the coefﬁcient to the drainage area. To make a comparison to these data, in Fig. 12.8 (below) we have plotted simulated sediment discharge per unit width (km s1 m1) to water discharge (m3 s1) for three periods that are studied in more detail in the paper presented. These correspond to medium (50–100 years cal. BP), high (100–150 years cal. BP) and very high (150–200 years cal. BP) ﬂow conditions. They have deliberately been plotted with the same scales and dimensions to Fig. 7 in Barry et al. (2004)’s paper to allow a direct comparison to their data from the Boise River. In order to generate these plots we have converted daily sediment yields to km m1 s1. As the data generated by CAESAR is the total sediment leaving the right-hand edge of the modelled DEM there is no ﬁxed width across which sediment output is measured. As such the sediment yield from the catchment may be exit from one cell or from many (often depending upon the size of the event). Therefore, to make this comparison we have assumed a constant channel width of 100 m (two cells). This assumption may have an important effect on the comparison and this will be discussed later in this text.

1

1

1

0.1

0.1

0.1

0.01

0.01

0.01 y = 2E-09x 4.2803 R2 = 0.2525

y = 1E-06x2. 8015 R2 = 0.3749

y = 0. 0005x 1.5341 R2 = 0.1061

0. 001

0. 001

0. 001

0.0001

0.0001

0.0001

0.00001

0.00001 10

100

1000

0.00001 10

100

1000

10

100

1000

Figure 12.8. Sediment discharge power relationships for 50–100 years (left), 100–150 years (centre) and 150–200 years (right) cal. BP from the simulations presented. These results are plotted as per Fig. 7 in Barry et al. (2004) with discharge in m3 s1 on the x axis, and total transport per unit width in kg s1 m1 on the y axis. The grey line corresponds to the relationship identiﬁed for the Boise River by Barry et al. (2004).

Non-stationarity of basin scale sediment delivery

333

The results from CAESAR compare favourably with those from Barry et al. (2004) and the coefﬁcients and exponents shown in Fig. 12.8 ﬁt within the ranges found in their work. This is encouraging, but possibly not too surprising as the range of exponent and coefﬁcient values found by Barry et al. (2004) is wide. For a more detailed comparison, the grey line in Fig. 12.8 corresponds with the data Barry et al. (2004) present for the Boise River. This shows a similar exponent (angle of the line) to the centre 100–150 year cal. BP CAESAR example but a substantially different offset. Examining the data presented by Barry et al. (2004) the modelled Swale is considerably smaller than the Boise River (ca. 380 to 2500 km2) and should therefore have a much larger coefﬁcient, raising the position of the line. The angle of the lines from the CAESAR simulations also changes with a shallower line in the wetter example (150–200 cal. BP) and a steeper gradient in the drier run (50–100 cal. BP). Barry et al. (2004) suggest that the exponent (controlling the angle of the curve) increases when there is a better formed (more distinct) armour layer and reduces when there is less difference between surface and subsurface grain sizes. This would concur with these simulations, where during high ﬂow (wetter) conditions the armour layer is less well established, but during lower ﬂow examples there is more opportunity for an armour to develop. As such, the dynamics of the simulations we have presented ﬁt well with the theory used to explain the ﬁndings of Barry et al. (2004). However, as previously mentioned, the assumed channel width can have an important effect on these graphs, and the angle of the relationship in the simulation results could easily be manipulated by changing the width. There are also many low value data points from our data, which is possibly as the numerical model will record very small sediment yields that might not register in ﬁeld sampling. In summary, there are some differences in the sediment discharge data that we have simulated and that measured by Barry et al. (2004). However, the modelled data ﬁts comfortably within the range of coefﬁcient and exponent values measured by Barry et al. (2004) and more importantly the dynamics of how that exponent changes corresponds with the ﬁeld data. As such we conclude that this comparison provides an encouragingly good but not conclusive indication of model performance. Reference Barry, J.J., Bufﬁngton, J.M., King, J.G., 2004. A general power equation for predicting bed load transport rates in gravel bed rivers. Water Resour. Res. 40, W10401.

Discussion by Bob Mussetter The model provides some very interesting results regarding the long-term variability in sediment loads. Based on the presentation at the conference, the authors appear to interpret these results to mean that available sediment-transport relationships are ﬂawed because they do not account for changes in sediment supply and climate. In reality, the equations used in the model predict the transport capacity (or potential) for the concurrent bed material and hydraulic conditions, which depend on the

Non-stationarity of basin scale sediment delivery

333

The results from CAESAR compare favourably with those from Barry et al. (2004) and the coefﬁcients and exponents shown in Fig. 12.8 ﬁt within the ranges found in their work. This is encouraging, but possibly not too surprising as the range of exponent and coefﬁcient values found by Barry et al. (2004) is wide. For a more detailed comparison, the grey line in Fig. 12.8 corresponds with the data Barry et al. (2004) present for the Boise River. This shows a similar exponent (angle of the line) to the centre 100–150 year cal. BP CAESAR example but a substantially different offset. Examining the data presented by Barry et al. (2004) the modelled Swale is considerably smaller than the Boise River (ca. 380 to 2500 km2) and should therefore have a much larger coefﬁcient, raising the position of the line. The angle of the lines from the CAESAR simulations also changes with a shallower line in the wetter example (150–200 cal. BP) and a steeper gradient in the drier run (50–100 cal. BP). Barry et al. (2004) suggest that the exponent (controlling the angle of the curve) increases when there is a better formed (more distinct) armour layer and reduces when there is less difference between surface and subsurface grain sizes. This would concur with these simulations, where during high ﬂow (wetter) conditions the armour layer is less well established, but during lower ﬂow examples there is more opportunity for an armour to develop. As such, the dynamics of the simulations we have presented ﬁt well with the theory used to explain the ﬁndings of Barry et al. (2004). However, as previously mentioned, the assumed channel width can have an important effect on these graphs, and the angle of the relationship in the simulation results could easily be manipulated by changing the width. There are also many low value data points from our data, which is possibly as the numerical model will record very small sediment yields that might not register in ﬁeld sampling. In summary, there are some differences in the sediment discharge data that we have simulated and that measured by Barry et al. (2004). However, the modelled data ﬁts comfortably within the range of coefﬁcient and exponent values measured by Barry et al. (2004) and more importantly the dynamics of how that exponent changes corresponds with the ﬁeld data. As such we conclude that this comparison provides an encouragingly good but not conclusive indication of model performance. Reference Barry, J.J., Bufﬁngton, J.M., King, J.G., 2004. A general power equation for predicting bed load transport rates in gravel bed rivers. Water Resour. Res. 40, W10401.

Discussion by Bob Mussetter The model provides some very interesting results regarding the long-term variability in sediment loads. Based on the presentation at the conference, the authors appear to interpret these results to mean that available sediment-transport relationships are ﬂawed because they do not account for changes in sediment supply and climate. In reality, the equations used in the model predict the transport capacity (or potential) for the concurrent bed material and hydraulic conditions, which depend on the

334

T.J. Coulthard, J. Lewin, M.G. Macklin

previous supply and hydraulic conditions. As a result, the model does not illustrate a ﬂaw in the available transport equations, but rather likely helps explain at least part of the reason for the large variability that is observed in bed-material load measurements. Using bed-material transport relationships in this manner also does not address the dynamics of the wash (or ﬁne sediment) load, which is nearly always supply-limited, but an important part of the sediment dynamics of the watershed scale. Discussion by Colin D. Rennie Coulthard et al. and Pizzuto et al. have tackled the important issue of the impact of climate change on river channels. River managers are increasingly concerned with climate change, and require informed estimates of likely changes in river channels (e.g., Ashmore and Church, 2001). Coulthard et al. employed a sophisticated cellular landscape evolution model (CAESAR), which calculates spatiotemporally distributed sediment delivery and transport, to understand channel changes in time with and without climate change. Pizzuto et al. employed a relatively simple sediment transport model for the same ends. They argued that a simple model was justiﬁed because data and theory were not available for a more complex model. Coulthard et al. describe well the limitations of empirical approaches to understanding climate change impacts on rivers. The existing record of ﬂuvial measurements is too short to demonstrate morphological change due to climate change, and the record is convolved with channel changes due to changing land-use. The paleolimnological record from stratigraphy is too coarse and uncertain to demonstrate climate change impacts. As a consequence, numerical modelling has been employed to evaluate the inﬂuence of climate change on channels. However, numerical models require validation and calibration with measured data if they are to generate reliable predictions. The data problems that limit empirical approaches also limit validation of numerical models, as acknowledged by Coulthard et al., which renders suspect the numerical model outcomes. The model results are highly uncertain, even with respect to observed trends, thus models may not even serve as exploratory tools. This is a difﬁcult dilemma, which the gravel-bed rivers community should begin to address. Gordon Grant suggested hindcasting numerical models to compare with the sedimentological record, which may be a good ﬁrst step.

Reference Ashmore, P., Church, M., 2001. The impact of climate change on rivers and river processes in Canada, Geol. Surv. of Canada, Bulletin 555, 58pp.

Reply by the authors We agree with both sets of comments. This work was not designed to out of hand discount sediment-transport formulae, but to point out that there were some

334

T.J. Coulthard, J. Lewin, M.G. Macklin

previous supply and hydraulic conditions. As a result, the model does not illustrate a ﬂaw in the available transport equations, but rather likely helps explain at least part of the reason for the large variability that is observed in bed-material load measurements. Using bed-material transport relationships in this manner also does not address the dynamics of the wash (or ﬁne sediment) load, which is nearly always supply-limited, but an important part of the sediment dynamics of the watershed scale. Discussion by Colin D. Rennie Coulthard et al. and Pizzuto et al. have tackled the important issue of the impact of climate change on river channels. River managers are increasingly concerned with climate change, and require informed estimates of likely changes in river channels (e.g., Ashmore and Church, 2001). Coulthard et al. employed a sophisticated cellular landscape evolution model (CAESAR), which calculates spatiotemporally distributed sediment delivery and transport, to understand channel changes in time with and without climate change. Pizzuto et al. employed a relatively simple sediment transport model for the same ends. They argued that a simple model was justiﬁed because data and theory were not available for a more complex model. Coulthard et al. describe well the limitations of empirical approaches to understanding climate change impacts on rivers. The existing record of ﬂuvial measurements is too short to demonstrate morphological change due to climate change, and the record is convolved with channel changes due to changing land-use. The paleolimnological record from stratigraphy is too coarse and uncertain to demonstrate climate change impacts. As a consequence, numerical modelling has been employed to evaluate the inﬂuence of climate change on channels. However, numerical models require validation and calibration with measured data if they are to generate reliable predictions. The data problems that limit empirical approaches also limit validation of numerical models, as acknowledged by Coulthard et al., which renders suspect the numerical model outcomes. The model results are highly uncertain, even with respect to observed trends, thus models may not even serve as exploratory tools. This is a difﬁcult dilemma, which the gravel-bed rivers community should begin to address. Gordon Grant suggested hindcasting numerical models to compare with the sedimentological record, which may be a good ﬁrst step.

Reference Ashmore, P., Church, M., 2001. The impact of climate change on rivers and river processes in Canada, Geol. Surv. of Canada, Bulletin 555, 58pp.

Reply by the authors We agree with both sets of comments. This work was not designed to out of hand discount sediment-transport formulae, but to point out that there were some

334

T.J. Coulthard, J. Lewin, M.G. Macklin

previous supply and hydraulic conditions. As a result, the model does not illustrate a ﬂaw in the available transport equations, but rather likely helps explain at least part of the reason for the large variability that is observed in bed-material load measurements. Using bed-material transport relationships in this manner also does not address the dynamics of the wash (or ﬁne sediment) load, which is nearly always supply-limited, but an important part of the sediment dynamics of the watershed scale. Discussion by Colin D. Rennie Coulthard et al. and Pizzuto et al. have tackled the important issue of the impact of climate change on river channels. River managers are increasingly concerned with climate change, and require informed estimates of likely changes in river channels (e.g., Ashmore and Church, 2001). Coulthard et al. employed a sophisticated cellular landscape evolution model (CAESAR), which calculates spatiotemporally distributed sediment delivery and transport, to understand channel changes in time with and without climate change. Pizzuto et al. employed a relatively simple sediment transport model for the same ends. They argued that a simple model was justiﬁed because data and theory were not available for a more complex model. Coulthard et al. describe well the limitations of empirical approaches to understanding climate change impacts on rivers. The existing record of ﬂuvial measurements is too short to demonstrate morphological change due to climate change, and the record is convolved with channel changes due to changing land-use. The paleolimnological record from stratigraphy is too coarse and uncertain to demonstrate climate change impacts. As a consequence, numerical modelling has been employed to evaluate the inﬂuence of climate change on channels. However, numerical models require validation and calibration with measured data if they are to generate reliable predictions. The data problems that limit empirical approaches also limit validation of numerical models, as acknowledged by Coulthard et al., which renders suspect the numerical model outcomes. The model results are highly uncertain, even with respect to observed trends, thus models may not even serve as exploratory tools. This is a difﬁcult dilemma, which the gravel-bed rivers community should begin to address. Gordon Grant suggested hindcasting numerical models to compare with the sedimentological record, which may be a good ﬁrst step.

Reference Ashmore, P., Church, M., 2001. The impact of climate change on rivers and river processes in Canada, Geol. Surv. of Canada, Bulletin 555, 58pp.

Reply by the authors We agree with both sets of comments. This work was not designed to out of hand discount sediment-transport formulae, but to point out that there were some

Non-stationarity of basin scale sediment delivery

335

signiﬁcant weaknesses and areas of uncertainty that may need to be addressed. Colin D. Rennie suggests that the high levels of uncertainty found in complex models such as CAESAR may well make them inappropriate to use even for exploratory or hypothesis testing tools. This is undoubtedly a concern, and an issue shared by climate modellers, but we wish to present these results and the model itself not as a ﬁnished article, but as a ﬁrst step towards developing better constrained more accurate numerical models. Even if the results of CAESAR are completely wrong, it has served an important duty by revealing uncertainties and raising debate in how we look at sediment transport and basin sediment delivery. Colin also raises an important ﬁnal point, with the possibility of using hind casting or retro-validation to simulate the past and compare the results to the present day or sedimentological record. This is a technique used by climate modellers and that has also been applied to CAESAR in two publications (Coulthard and Macklin, 2001; Coulthard et al., 2005) where they compare the long-term sediment yield to a sedimentary record.

References Coulthard, T.J., Lewin, J., Macklin, MG., 2005. Modelling differential and complex catchment response to environmental change. Geomorphology 69, 224–241. Coulthard, T.J., Macklin, M.G., 2001. How sensitive are river systems to climate and land-use changes? A model based evaluation. J. Q. Sci. 16 (4), 347–351.

Gravel-Bed Rivers VI: From Process Understanding to River Restoration H. Habersack, H. Pie´gay, M. Rinaldi, Editors r 2008 Elsevier B.V. All rights reserved.

337

13 Changes in basin-scale sediment supply and transfer in a rapidly transformed New Zealand landscape Mike Page, Mike Marden, Mio Kasai, Basil Gomez, Dave Peacock, Harley Betts, Thomas Parkner, Tomomi Marutani and Noel Trustrum

Abstract Society has an ever-increasing need to manage landscapes. To do this effectively requires improved understanding of the way landscapes behave, and the controls on that behaviour. This is certainly the case where sustainable resource use and hazard mitigation involve the management of the generation, transport, and storage of sediment. Landscapes are complex systems, consisting of a mosaic of landforms. At the broad regional level these landforms are arranged in characteristic patterns, reﬂecting environmental conditions and associated processes. At the catchment level, assemblages of landforms have a unique conﬁguration, forming an interactive functioning system through which water and sediment are passed. It is this unique, catchmentbased assemblage that we seek to manage. Controls on the way landscapes behave are numerous and operate at a variety of scales both spatial and temporal. The way these controls interact on a complex and unique arrangement of landforms is difﬁcult to predict. The East Coast of the North Island of New Zealand is a dynamic landscape. High natural erosion rates have been augmented by recent and rapid anthropogenic activity. Several studies in the Waiapu catchment, involving a range of spatial and temporal scales, are used here to illustrate the impact of natural and anthropogenic controls on basin-scale sediment supply and transfer. In this landscape, the use of vegetation, speciﬁcally targeted reforestation, is the most effective method of sediment management. This will be enhanced by improved understanding of stability thresholds and hill slope–channel connectivity.

E-mail address: [email protected] (M. Page) ISSN: 0928-2025

DOI: 10.1016/S0928-2025(07)11132-9

M. Page et al.

338 1.

Introduction

Landscapes are complex systems. They are the products of the interaction of a set of environmental controls. Overprinted onto a geomorphic template determined by geology, climate, and time, are the inﬂuences of numerous other landscape-forming controls. These include vegetation, human impacts and the nature and timing of so-called random events (e.g., climatic, tectonic, volcanic). The result is a mosaic of landforms that, because of their unique evolutionary histories, have different levels of sensitivity, resistance, and resilience to further environmental change or system perturbations, both natural and human-induced. The ability to predict and manage the input of sediment to rivers at the basin-scale is confounded by this complexity.

1.1.

Controls

Controls operate at different scales. At the broadest scale, controls imposed by tectonic and climatic setting and regional geology and vegetation set the boundary conditions for behaviour (Brooks and Brierley, 1997). The history of interaction between these controls has led to basin size and form, including landforms and their topology. These in turn inﬂuence hill slope–channel coupling, drainage pattern and density, channel gradient and morphology, and hence ultimately sediment supply and transport. The state or stage of evolution of the landscape will also inﬂuence sediment generation and movement, the spatial and temporal connectivity/disconnectivity of sediment transfer linkages, and residence time of sediment in storage sites. At the within-basin-scale, temporal and spatial sequencing of sediment inputs is also controlled by local geomorphology, human impacts (especially land use/landuse change), and local rainfall – which in the case of storm magnitude and frequency are quasi-random, both spatially and temporally (Table 13.1). These controls are, by and large, well known. However, the results and products of their interaction are difﬁcult to predict. Opportunities to gain such insights are often greatest in dynamic landscapes where response to changing environmental conditions is both rapid and pronounced. Controls on the patterns and rates of sediment delivery in a range of environmental settings have been highlighted and compared in a series of case studies from around the Paciﬁc Rim (Marutani et al., 2001). Recent studies of land-use change, sediment production, and channel response include: Knighton (1989), Brooks and Brierley (1997), Kondolf et al. (2002), Lie´bault et al. (2002), Owens and Walling (2002), Surian and Rinaldi (2004), and Lie´bault et al. (2005).

1.2.

Sensitivity

The concept of landscape sensitivity, ﬁrst promoted by Brunsden and Thornes (1979), provides a framework to evaluate landscape behaviour and change, and states that ‘‘a given change in the controls of a system or the forces applied to the system will likely produce a sensible, recognisable, sustained but complex response’’.

Changes in basin-scale sediment supply and transfer Table 13.1.

339

Controls on erosion/sediment supply.

Geologic controls

Climatic controls

Geomorphic controls

Anthropogenic controls

State of landscape evolution

Rainfall regime

Slope/ morphology

Geologic structure

Rainfall intensity

Weathering rates

Tectonism Base-level change Earthquakes

Storm magnitude–frequency

Channel network

Land use/ vegetation cover Land management history Channel modiﬁcation

Volcanism

Slope–channel connectivity

Sediment extraction, retention, augmentation

Rock type

The sensitivity of landscapes to erosion and the supply of sediment to rivers (and therefore river processes and morphology) are likewise conditioned by the interaction between these environmental controls. Landscape sensitivity has now become a widely recognised geomorphological concept for addressing issues of landscape change (Thomas and Allison, 1993; Brunsden, 2001; Thomas, 2001). Sensitivity to change varies spatially and temporally, and is inﬂuenced by such factors as sediment exhaustion and recovery, magnitude–frequency relationships, temporal and spatial sequencing, sediment delivery, storage and lags, and feedback mechanisms. Landscapes and their components may be sensitive to small changes or only to large disturbances. The response may also be small, or large and catastrophic, and range from gradual to immediate. Similarly, landscapes may exhibit different rates of recovery. This paper provides examples from a landscape where response to environmental change is rapid.

1.3.

Thresholds

The sensitivity of the landscape to erosion is conditioned by many factors that set the threshold for movement. Recognising stability thresholds of change, or points at which the system is unable to resist forces imposed by environmental conditions (Schumm, 1977), and managing the landscape to avoid exceeding these thresholds, or to exceed them in a managed, controlled fashion is a pragmatic way of dealing with what is a complex system. General patterns of response to disturbance are landscape-speciﬁc, determined by such characteristics as geology, relief/slope, channel form and network, and vegetation. However, variability in basin conﬁguration and the complex interactions

M. Page et al.

340

between landforms means the timing and rate of response is likely to be basinspeciﬁc. Of particular importance to sediment ﬂux are the thresholds for erosion, and the connectivity between hill slopes and channels, both of which are constantly adjusting to within-basin disturbances. Thresholds for erosion may change due to ‘‘external’’ factors such as a change in vegetation, or in response to a change in the magnitude and frequency of rainfall events (Wolman and Gerson, 1978). Thresholds may also change due to ‘‘internal’’ geomorphic factors such as the incremental accumulation of regolith on a slope, or the over-steepening of hill slopes through stream incision, leading to a lowering of base level. Conversely, if regolith is eroded at a greater rate than it can reform, then the amount of sediment available is reduced and is now conﬁned to more resistant sites, raising the threshold required to trigger future movement (Crozier, 1996). Furthermore, landscape components may not behave independently, and exceeding a threshold in one component may cause instability elsewhere.

1.4.

Connectivity

Once an erosion threshold has been exceeded, leading to the mobilisation of sediment, the delivery and transfer of that sediment is a function not only of hill slope and channel processes, but also of the connectivity of various landscape components. Just as thresholds change over time, so do coupling relationships. The impact of anthropogenic activity has generally, but not necessarily, been to increase connectivity. Much research has been directed at components of the landscape such as small catchments, individual landforms or hill slope elements (Madej, 1995; Nolan and Marron, 1995; Fryirs and Brierley, 1999; Harvey, 2001; Harvey, 2002). However, basin-wide linkages of these components are more difﬁcult to address (Wainwright et al., 2002). Improved understanding of thresholds of stability, and of connectivity, both between hill slopes and channels and between channel reaches, offer the best opportunity for successfully-targeted management.

2.

Area description

For the last two decades, research on the East Coast of the North Island of New Zealand (Fig. 13.1) has been focused on discriminating tectonic, climatic, and anthropogenic controls on landscape behaviour and change, including sediment supply and transfer. The region is located in a dynamic tectonic and climatic setting. In combination with lithology and topography, this has resulted in a landscape that has high natural erosion rates and is sensitive to changes in environmental conditions. Evidence of rapid landscape change is preserved in high-resolution geomorphic and sedimentary records. Overprinted onto natural ‘‘background’’ rates of change is the impact of recent, rapid anthropogenic activity, making this an attractive environment in which to study controls on sediment generation and transfer. With an annual average suspended sediment yield of 69 Mt/yr the region accounts for 33% of the

Changes in basin-scale sediment supply and transfer

341

Figure 13.1. Location of Waiapu catchment, and Raparapaririki, Mangawhairiki, and Weraamaia study catchments.

total New Zealand yield of 209 Mt/yr from only 2.5% of the land area (Hicks et al., 2002).

2.1.

Waiapu catchment

The 1734-km2 Waiapu catchment is located in the northeast of the region (Fig. 13.1). It forms part of a tectonically active and structurally complex zone referred to as the East Coast Allochthon, a series of thrust sheets, which was emplaced in the Early Miocene as a result of southwest to SSW-directed subduction of the Paciﬁc Plate beneath the Australian Plate (Mazengarb and Speden, 2000). Uplift rates are of the order of 1–4 mm/yr, with rates highest in headwaters draining the Raukumara Peninsula (Ota et al., 1992). There is an associated high incidence of earthquakes. The ﬂuvial network is continuing to adjust to this base-level change through incision. This adjustment is propagating upstream at varying rates depending on such factors as drainage area, lithology, and initial channel conditions. Incision has yet to reach some upland areas of more resistant and ﬁne-grained rocks, and the topography and channel network in these areas remain unadjusted. In parts of the catchment rapid rates of stream incision have resulted in several ﬂights of late Quaternary terraces. A complex pattern and wide range of rock types are present in the Waiapu catchment, the major ones being Tertiary and Quaternary mudstone and sandstone, and indurated sandstones and mudstones (termed ‘‘argillite and greywacke’’) of Mesozoic age, which form the Raukumara Range in the west. Large-scale folding, faulting and associated numerous shear zones, have made many of the rock types susceptible to

342

M. Page et al.

gully and deep-seated earthﬂow erosion. The term gully is here used to include both ﬂuvial incision- and mass movement-dominated gully forms. Most of the unstable rocks belong to the East Coast Allochthon. Major volcanic eruptions from Taupo and Rotorua volcanic centres during the last 100,000 years have deposited a number of airfall tephras over the catchment. However, due to high natural ‘‘geologic’’ rates of erosion and rapid removal of indigenous vegetation cover in the last 100 years, the tephra cover is now often thin, patchy, or has been completely removed. The presence or absence of tephras can therefore be used to indicate the level of stability of different landforms. Drainage runs in an east and northeast direction towards the Paciﬁc Ocean. Drainage patterns are random where bedrock is of uniform poor strength, such as crushed and sheared rock, but strongly controlled where rocks are indurated and bedded. In the headwaters of the Waiapu catchment rainfall averages between 2400 and 4000 mm/yr, and erosion-generating storms have a recurrence interval of 2.6 years (Hicks, 1995). New Zealand was the last major land mass to be inhabited by humans. Maori settlement in the Waiapu catchment can be traced back 600 years. When Europeans arrived in 1840 the catchment was largely covered in indigenous podocarpbroadleaved forest and beech forest (80%). Deforestation and the establishment of pastoral farming by European settlers began in earnest about 1890, and continued until 1920. A consequence of deforestation has been a dramatic increase in erosion and sediment transfer, in some hill country areas by at least an order of magnitude, indicating that land-use-driven changes in vegetation cover override other controls on erosion. The pulse of sediment generated by this change from forest to pasture is currently being redistributed across the landscape and in the marine environment. Reforestation for erosion control began in the Waiapu catchment in the late 1960s, and today exotic forest occupies 26% of the catchment. In many areas the deforestation/reforestation cycle has occurred in the space of 50–100 years. The East Coast Forestry Project is a government-operated scheme that has the aim of ‘‘controlling erosion on severely eroding target land’’ by 2020, mainly through reforestation (Ministry of Forestry, 1994; Phillips and Marden, 2004). At current planting rates, a further 15–20% of the Waiapu catchment may be reforested by 2020. The Waiapu catchment now has an annual average suspended sediment yield of 35 Mt/yr or 20,520 t/km2/yr (Hicks et al., 2000). This equates to 0.2% of the global yield and ranks among the highest recorded anywhere in the world (Walling and Webb, 1996). The majority of this sediment is supplied from large gullies (more correctly termed gully-mass movement complexes). Gullies have developed in crushed and sheared Cretaceous-aged sedimentary rocks that occupy more than 20% of the Waiapu catchment, while landslides occur on Tertiary-aged mudstones and sandstones. The Cretaceous rocks generate the majority of the coarse (gravel to boulder) bed material. However, the high degree of deformation undergone by these rocks means they are often weak, and subsequently breakdown during in-channel storage and transport, resulting in high-suspended sediment loads. The sediment yield from the Waiapu catchment is about 2.5 times that of its southern neighbour the Waipaoa catchment, which has a suspended sediment yield of 15 Mt/yr (6800 t/km2/yr).

Changes in basin-scale sediment supply and transfer

343

In the Waipaoa catchment gullies contribute between 50 and 60% of the annual suspended sediment yield, and 15% is derived from landslides. For the Waipaoa River, bed load is estimated to be 1% of the suspended load. Both catchments are of similar size and have a similar geologic and climatic setting and deforestation history. However, gully contribution to sediment yield is greater in the Waiapu catchment which has twice the area of gully-prone terrain, and 900 active gullies (Marden, 2003) compared with 420 in the Waipaoa catchment (Marden et al., 2005). The 900 gullies have a combined area of 4028 ha, and the catchment area associated with these gullies accounts for 25.5% of the Waiapu catchment. The difference between the 30,000 t/km2/yr collectively generated by these gullies and the catchment suspended sediment yield of 20,520 t/km2/yr (derived from all erosion processes) indicates signiﬁcant volumes of sediment are continuing to be stored in the channel network. For large magnitude events in gullied upper tributaries of the Waipaoa catchment, sediment delivery ratios (SDR) of 40.90 have been estimated (Marutani et al., 1999; Kasai et al., 2001). This channel network will be a major source of sediment for many decades (Gomez et al., 2003). In areas dominated by erodible Cretaceous rocks the volume of sediment generated following deforestation has led to rapid aggradation, which has overwhelmed the background, base-level-controlled incision of the ﬂuvial network. The Waiapu River is now typically braided and dominated by deep gravel beds in its lower (50 km) reaches. In mid-to-upper reaches, in areas of hard Tertiary mudstone and sandstone rocks, the river is incised and has a more single-threaded and meandering form. Aggradation of rivers and streams is a major threat to the Waiapu ﬂoodplain (which is unprotected by stopbanks/levees), roads, and bridges. The town of Ruatorea and several small community settlements are located near or adjacent to the ﬂoodplain. Aggradation has also impacted adversely on river health and ecology. Aggradation is particularly evident in the Tapuaeroa (Fig. 13.2) and Mangaoporo, and some lower parts of the Mata, Lower Mata, and Ihungia subcatchments. Channel widening in these areas is greatly affecting the ﬂoodplain, mainly through inundation and bank erosion, and many meander bends are undermining adjacent hillsides. However, with few large storms since Cyclone Bola in 1988, and continued reforestation, a reduction in sediment supply from hill slopes has resulted in

Figure 13.2. Looking up the bed of the Tapuaeroa River towards Mt. Wharekia. Note large gully complex centre left.

M. Page et al.

344

degradation in many upper tributary reaches. Nevertheless, vast quantities of sediment are stored in the stream channel network as bedload, ensuring that as this phase of down cutting progresses, the sediment supply to the ﬂoodplain will remain high. The population of the catchment is largely indigenous Maori (90%), and deforestation, subsequent erosion and sedimentation, and the associated loss of ﬂora and fauna, have had an enormous impact on cultural values and the use of traditional resources. However, a growing drive for self-determination is leading to the establishment of innovative land-based industries and the desire for ecosystem restoration (Page et al., 2001; Harmsworth et al., 2002; Harmsworth and Raynor, 2004). Rapid deforestation and equally rapid reforestation, in a catchment where process rates are also high, provides a rare opportunity to study the geomorphic response to these land-use changes, including the processes that redistribute sediment. Key questions for affected communities are at what rate, over what time scale, and to what level will the landscape recover and how will this inﬂuence future land-use options (Page et al., 2000)? Several studies in the Waiapu catchment, involving a range of spatial and temporal scales, are used to illustrate the impact of natural and anthropogenic controls on basin-scale sediment supply and transfer. They range from studies of the morphological development and sediment supply of an individual gully in response to rainfalls during the course of 1 year period, to decadal changes to sediment supply and transfer in a small catchment, in response to changing land use/vegetation, to the impact of a high magnitude rainstorm on naturally forested slopes in a medium-sized catchment.

3. 3.1.

Results Gully erosion and reforestation in the Waiapu catchment

In 1997 there were 901 active gullies in the Waiapu catchment (Marden, 2003), and 556 that had been stabilised (Marden, personal communications). The majority of gullies developed within the ﬁrst two to three decades after deforestation. Active gullies occur in seven major lithological formations. The sediments that enter streams from these different materials have characteristic morphology, composition, particle size, hardness, and weathering rates, which strongly inﬂuences sediment storage and rate of sediment transfer through the channel network. Gully development is a highly dynamic process controlled by a combination of vegetation and the timing of vegetation change, together with rainfall magnitude and frequency. Gullies switch on and off, some stabilise naturally through scrub reversion, others through reforestation. The result is gullies at different stages of development. A study of the effectiveness of reforestation in controlling gully development and stabilising gullies in Cretaceous rocks found key determinants are the size and shape of gullies at time of planting (Marden et al., 2005). Probabilities of success are: 480% for gullies o1 ha in area, 60% for gullies 1–5 ha in area, 50% for gullies of 5 ha, and little chance of success for gullies 410 ha in area. Linear gullies are more

Changes in basin-scale sediment supply and transfer

345

likely to be stabilised than amphitheatre-shaped gullies. About 100 of the 901 active gullies (11%) are considered too large to be stabilised by revegetation techniques. The majority of these occur in areas of indigenous forest in the steep headwaters of the Raukumara Range.

3.2. Decadal changes in sediment supply and transfer in response to changing land use in the Weraamaia catchment The Weraamaia stream is a 4.8-km2 tributary of the Mangaoporo River, which in turn drains into the Waiapu River (Fig. 13.3). The Weraamaia catchment is representative of the gully-prone hill country that supplies much of the sediment to the Waiapu River. It was chosen for study because it has subcatchments in indigenous forest, exotic forest, and pasture that provided the opportunity to study land-use controls on gully development, and stream channel morphology and behaviour through changes in hydrology and sediment supply. The majority of the catchment is underlain by mudstone of the Whangai Formation, while indurated, sandstonedominated alternating sandstone and mudstone of the Tapuwaeroa Formation underlie the northernmost 10% of the catchment (Mazengarb and Speden, 2000). By the early 1920s much of the Mangaoporo catchment, and all but a 0.55 km2 third-order subcatchment of the Weraamaia catchment, had been cleared of forest by a combination of burning and logging. This is now the largest remaining gully-prone, argillite drainage basin in the district with a complete original forest cover. Soon after conversion to pasture, fern, scrub, and weed species became established in the Weraamaia catchment, followed by periodic scrub clearance and re-establishment. By 1939, the date of the ﬁrst aerial photographs, scattered scrub occurred only in the

Figure 13.3. Weraamaia catchment. (a) 1957 (NZ Aerial Mapping Ltd., Hastings). Indigenous forest in northeast, pasture in southeast, and pasture reverting to scrub in west of catchment. (b) 1988 (Aerial Surveys Ltd., Nelson). Exotic forest (8 years old) in west of catchment. Arrow indicates a debris ﬂow generated from a gully complex during Cyclone Bola.

346

M. Page et al.

head of the catchment. By 1971 heavy scrub covered the head of the catchment and parts of the central and lower catchment. The western side of the catchment was planted in exotic forest (Pinus radiata) in 1979–1980, and since 1996 areas of pasture on the south-eastern side have also been successively planted in exotic forest. Woody vegetation (indigenous forest, exotic forest, scrub) has increased from 38% in 1939 to 45% in 1984 to 85% in 1997. The chronology of gully development has been documented by Parkner et al. (2006) using sequential aerial photographs. Patterns of sediment ﬂux and channel morphology changes have been identiﬁed by Kasai et al. (2005) from analysis of sequential aerial photographs, dating of terraces surfaces, and valley ﬂoor crosssections and longitudinal proﬁles recorded using a laser distance measure and laser level. Annual sediment yields from gullies were derived using empirical equations describing volume–area relationships (Betts and DeRose, 1999). By 1939 almost all deep-seated gullies and gully complexes had already been initiated, all in areas that had been cleared of forest, and several (5) had already stabilised. Gullies and channels were well coupled. Shallow landslides were also common. Channels were aggrading and widening. Between 6.8 and 18.7 m above the contemporary (2002) channel, they were higher than at any other date. Much of this aggradation was due to a high magnitude storm event in 1938, estimated to be 4700 mm in 30 h. A number of rainfall events before 1953 triggered further erosion, so that by 1957 a maximum number of 19 gullies (20.9 ha) and four gully complexes (15.02 ha) were present. Thereafter there was a reduction in the number and size of features, until by 1988 there were only six gullies and three gully complexes active. Channels narrowed and incised in response to the decrease in sediment yield from hill slopes. Most features were then reactivated by Cyclone Bola in 1988 (4600 mm in 4 days). Shallow landslides were also common, even in the forested subcatchments. The main sediment impact however was a debris ﬂow (Fig. 13.3) derived from a gully complex that formed a debris fan that temporarily blocked the main channel. The volume of this fan was 96,500 m3 (Kasai et al., 2005). Channels aggraded again, especially upstream from the fan. However, downstream bed levels were 7–8 m below those in 1938. There then followed a further reduction in gully activity, until by 2003 there were only two gullies (1.8 ha) and three gully complexes (1.75 ha) active. The majority of the sediment in this period was supplied from the debris ﬂow and gullies in areas of pasture. A large amount of fan material was removed, followed by incision, both upstream and downstream. Channels narrowed and incised to their lowest levels in the study period, which further decoupled hill slopes from channels. Narrowing and incision were most rapid in landslide-dominated, forested catchments, in response to rapid revegetation of scars. In places in the lower reaches of the main channel, incision has now reached bedrock, and rates of incision and narrowing have slowed. Total catchment sediment yield for four periods, and under a hypothetical complete forest cover, is given in Table 13.2. Yields from gullies and gully complexes have been reducing since 1939 as the area of the catchment in woody vegetation increases. Yield from landslides has followed a similar trend, punctuated by a temporary increase following Cyclone Bola. Yield from stored channel sediment has also been reducing, except for a large increase following Cyclone Bola in response to both the

Changes in basin-scale sediment supply and transfer Table 13.2. Period

347

Changes to sediment ﬂux in the Weraamaia catchment. 1939–1950

Erosion rate (m3/km2/yr) Gullies 6,825 Landslides 18,156 Channels 20,323 Catchment 45,304 sediment yield (m3/km2/yr)

1950–1984

1988–1991a

1991–2001

Hypothetical forest

3,456 10,569 9,469 23,494

3,171 18,846 28,902 50,919

1,167 723 4,612 6,502

0 667 2,252 2,919

Source: Derived from Kasai et al. (2005). a Erosion rate does not include debris ﬂow that occurred during Cyclone Bola.

increase in landslides and the large debris ﬂow (Fig. 13.3). This resulted in the immediate post-Bola catchment yield exceeding the post-1939 yield. Between 1939 and 1950 the sediment supplied to the channel from hill slopes (from gullies and landslides) exceeded sediment removed from channels (net gain); however, by 1991–2001 more sediment was removed from channels than was supplied to channels from hill slopes (net loss). The proportion of catchment sediment yield derived from channel beds was 45% between 1939 and 1950. This reduced to 40% between 1950 and 1984, and has increased since Cyclone Bola. Between 1988 and 1991 channel beds acccounted for 57% of total catchment yield, increasing to 71% between 1991 and 2001, and is approaching the ﬁgure of 77% predicted under a hypothetical complete forest cover. Parkner et al. (2006) also identiﬁed topographic thresholds for gully initiation based on catchment area and slope, which can be used to aid identiﬁcation of gullyprone areas under different vegetation (Patton and Schumm, 1975). Given that 490% of the Weraamaia catchment is now covered in exotic or indigenous forest or scrub, and the threshold for gully initiation or reactivation has risen, it is expected that future sediment supply, during even high magnitude events, will be considerably reduced. Following the present trajectory, remaining gullies will probably stabilise in a few decades (30 years), while headwater channels will continue to supply sediment for many decades. A major conclusion of these studies is that in this landscape, response to land-use/ vegetation change – both deforestation and reforestation – is very rapid. Just as deforestation induced rapid gully initiation and associated channel aggradation, reforestation has led to rapid degradation (Madej and Ozaki, 1996; Lie´bault et al., 2002). Gully development and behaviour is a highly dynamic phenomenon controlled by a combination of vegetation and the timing of vegetation change, together with rainfall magnitude and frequency. Development is not necessarily linear and may involve phases of expansion and inactivity. Gullies switch on and off, some stabilise naturally through scrub reversion, others through reforestation. The development and distribution of gullies since deforestation has dramatically modiﬁed the pattern and rate of sediment delivery to the valley ﬂoor and the processing of sediment through the channel network.

348

M. Page et al.

3.3. Raparapaririki catchment – an example of the impact of a large magnitude rainfall event on naturally forested slopes Although many rivers throughout the region have been aggrading as a result of deforestation, some headwater streams, where indigenous forest remains, have been stable or degrading. One such stream was the Raparapaririki, a tributary of the Tapuaeroa River, which is a major branch of the Waiapu River (Fig. 13.1). It drains a 35-km2 catchment, the upper 74% of which retains its indigenous forest cover. The remainder of the catchment, originally cleared for pasture in the late 1800s and early 1900s, is now in exotic forest. The exotic forest was planted in the early 1970s (true right bank) and in 2000 (true left bank) to control numerous linear gullies that had developed. The indigenous forest area is underlain by alternating sandstone and mudstone of the Tikihore and Tapuwaeroa Formations, and the remainder by Mokoiwi Formation sandstone and mudstone (Mazengarb and Speden, 2000). In March 1988 Cyclone Bola, the largest magnitude storm event on record in the catchment, caused extreme erosion within the indigenous forest. Subsequent concerns by the district council over the impact of channel aggradation on the road bridge across the Raparapaririki stream, and bridges across two other streams, led to a report on bed level changes and sediment sources within these catchments (Peacock and Marden, 2004). Nine cross-sections extending 3.4 km up the Raparapaririki stream have been surveyed regularly between 1974 and 2004 (Fig. 13.4). They show that before Cyclone Bola the stream was degrading at most cross-sections, leading to the exposure of supports on the road bridge (670 m upstream). At this time the channel was a steep bouldery mountain torrent with trout present. Rainfall during the 4 days of Cyclone Bola (extrapolated from nearest rain gauges), was likely to have been between 700 and 900 mm. In the headwaters of the catchment there was a 350% increase in the area of landslides and associated gullying (Lie´bault et al., 2005). In excess of 87% of the eroded area in the catchment is now within the area of indigenous forest (Fig. 13.4). The result was an immediate supply of gravel into the stream network due to the strong connectivity between erosion scars and channels. In places previously stable, vegetated terraces were buried by sediment. At the most upstream cross-section the streambed has aggraded 33.5 m since 1987 (last pre-Bola survey), at an average of 1.97 m/yr. A survey in October 1994 showed that the streambed at the bridge had aggraded 4.5 m since 1987, and by June 1996 had reached the top of the deck, when a new bridge was constructed 500 m downstream. Immediately downstream of the present bridge the bed has aggraded 7.2 m since 1987, and the bed width has increased from 134 to 191 m. In the 6 years following Cyclone Bola, 4.4 106 m3 of sediment accumulated in the 3.3-km-long reach upstream from the conﬂuence with the Tapuaeroa River, and a further 2.6 106 m3 of gravel accumulated in the succeeding 9-year period (Lie´bault et al., 2005). At the junction of the two rivers, aggradation in the Tapuaeroa River is half that of the Raparapaririki stream, leading to a ‘‘perching’’ of the Raparapaririki streambed. Following this prodigious aggradation there has been net incision and narrowing of the channel upstream of the 6-km cross-section (within a cut and ﬁll sequence), while downstream of this point the channel is still aggrading, with a ‘‘wave’’ of

Changes in basin-scale sediment supply and transfer

349

Figure 13.4. Comparison of SPOT 1 imagery obtained in 1986 (a) and SPOT 3 imagery obtained in 1996 (b) shows the magnitude of the storm-related deposition in Raparapaririki Stream following Cyclone Bola, as indexed by the increase in bed width (in black on the images). Location of the nine surveyed cross-sections are marked by solid dots (a). Dashed box delimits the area covered by the 1997 aerial photograph (c), which shows landslide scars and related erosion. Gully complex in the Mangawhairiki catchment (d). Note tension cracks and mass movement features around the periphery of the complex.

sediment passing along the streambed towards the new bridge. Assuming no major storm events, future prospects are for a reduced but continuing supply of sediment especially from gullies, with natural revegetation conﬁned to shallow landslide scars. The ‘‘wave’’ of sediment and continuing incision are expected to result in aggradation rates at the new bridge site of Z45 cm/yr for the next 5–10 years. Land-use-driven changes in vegetation cover exert the most important control on erosion in New Zealand. However, this example illustrates that, given certain physical conditions (in this case tectonic setting, erodible rock type, very steep slopes), forested landscapes are highly sensitive to erosion, and once a threshold is exceeded, these landscapes are capable of generating volumes of sediment of the same order of

350

M. Page et al.

magnitude as similar landscapes in pasture. The volumes of sediment generated are so great that they have a sudden and ongoing impact on the stream network. An important difference between natural and anthropogenic controls is that they tend to operate at different spatial and temporal scales, leading to different sequencing of sediment inputs. Storm events are episodic and ongoing and occur on temporally and spatially random bases. The result is that under natural conditions various parts of the catchment, through their different evolutionary and erosion histories, are at different stages of recovery and have different thresholds for further erosion (threshold variability). Land-use/vegetation changes, on the other hand, tend to occur over large areas over a short period of time. In the case of the Waiapu catchment, the conversion of forest to pasture in the space of 50–100 years has meant that erosion thresholds have been lowered and subsequently exceeded over much of the catchment, initiating a near uniformly timed phase of erosion, and subsequent trajectory of recovery. The result is a resetting of erosion thresholds, which reduces the threshold variability developed under natural conditions (threshold uniformity). However, in the case of the Waiapu catchment, reforestation of parts of the catchment over the last 35 years has again increased threshold variability by providing erosion protection to some areas. Future management for erosion control and sediment supply will be improved if thresholds for erosion are better understood. To achieve this it will be necessary to take into account not only the physical characteristics and function of the various landscape components, but also the storm history and land-use/vegetation history.

3.4. Gully development and sediment supply in response to rainfall in the Mangawhairiki catchment Sediment derived from gullies dominate sediment yield in the Waiapu catchment. The temporal behaviour of gully erosion – especially response to rainfall events – and processes of sediment generation, are important controls on the timing of sediment inputs. Studies of gullies in the Waipaoa catchment, using high-resolution digital elevation models derived from sequential aerial photography, have quantiﬁed geomorphic changes and long-term (decadal) sediment production and contribution to sediment yield (DeRose et al., 1998). The same technique has been employed to quantify short-term (13 months) geomorphic changes in, and sediment production from an 8.7-ha gully complex (Fig. 13.4d), located in the 320 ha Mangawhairiki catchment, which drains into the Tapuaeroa River, a tributary of the Waiapu River (Figs. 13.1 and 13.2) (Betts et al., 2003). The gully complex is in pasture, and occurs in alternating and deformed Cretaceous sandstone and mudstone of the Mokoiwi Formation (Mazengarb and Speden, 2000). Results showed sediment was generated by a complex process of mass movement (slumping and debris ﬂows) in response to over-steepening of gully sidewalls by channel incision. Mass movements accounted for almost 90% of the 520071700 m3 of sediment generated. A further 6707180 m3 was eroded from the associated 0.8-ha debris fan. In the early stages of development, gullies have a linear form as surface erosion and ﬂuvial incision dominate, but mass movement processes become increasingly

Changes in basin-scale sediment supply and transfer

351

important as a critical threshold of sidewall length and/or slope is passed, leading to a more amphitheatre gully shape. Surface erosion processes generally dominate during large storm events, whereas mass movement processes tend to become active several days later, due to the delay in ground water saturation, or following sustained wet periods. The dominance of mass movement over surface and ﬂuvial erosion processes means the relationship between individual rainfall events and gully response is likely to be weak, given that mass movement can continue for long periods in the absence of major storm events. Coupling between gully and channel varied with the magnitude and frequency of rainfall events and with gully development. Gully-channel coupling is strongest during the early, incision-dominated stages of development, weakening as gullies enlarge through mass movements and fans develop. Coupling also weakens during high magnitude events when large quantities of sediment are stored in fan deposits. These deposits are subsequently evacuated during small, frequent events. Again, targeting of mitigation techniques, especially reforestation, will be more effective with knowledge of gully complex development.

4.

Discussion and conclusions

The East Coast of the North Island of New Zealand is located in a dynamic tectonic and climatic setting. Overprinted onto high natural rates of erosion and sedimentation are the impacts of recent and extensive deforestation. Responses to environmental change are both rapid and pronounced. Studies by Betts et al. (2003), Peacock and Marden (2004), Kasai et al. (2005), and Parkner et al. (2006) in the Waiapu catchment, and others by DeRose et al. (1998), Marutani et al. (1999), Trustrum et al. (1999), Hicks et al. (2000), Kasai et al. (2001), Reid and Page (2002), Gomez et al. (2003), Lie´bault et al. (2005), Marden et al. (2005), and Owens et al. (2005), in the Waipaoa catchment, provide insights into controls on sediment generation and transfer, speciﬁcally the impact of deforestation and reforestation on gully and landslide behaviour. They indicate that land-use/vegetation change is the most effective method of inﬂuencing landscape behaviour at the basin-scale. The erosion response to both deforestation and reforestation has been rapid, in the order of 20–30 years. By 1997 there were 901 active gullies in the Waiapu catchment and 556 that had been stabilised. Gully location is controlled by lithology, degree of crushing, shearing and faulting, and by vegetation cover. Gully size is also controlled by lithology and deformation, and by age of gully (time since deforestation), rainfall regime including storm history (magnitude and frequency), and land use (regeneration or reforestation). Gullies are a chronic source of sediment. Once established they show little evidence of an erosion threshold. Thresholds for landsliding, on the other hand, are much higher, leading to often large, but less frequent inputs. During Cyclone Bola, landsliding provided 20.5 106 t or 64% of the estimated suspended sediment load of the Waipaoa River, compared with the annual average of 2.4 106 t (15%) (Page et al., 2000). Hicks et al. (2000) conﬁrm the importance of landsliding during high magnitude events in the Waipaoa catchment, by showing that in a subcatchment where gully processes dominate, half the

352

M. Page et al.

long-term average suspended sediment load is transported by events with return periods o1 year, while in a subcatchment dominated by landsliding half the long-term average suspended sediment load is transported by events with return periods 42 years. Nevertheless, rare, extreme rainstorms have a major inﬂuence on the landscape. Gullies can develop in indigenous forest given an event of sufﬁcient size, and the majority of the largest gullies now occur in indigenous forest. Reforestation is effective in stabilising gullies up to 10 ha in size, with probability of success greatest for small, linear gullies. Gullies will stabilise in decades (30 years). Changes in active gully complex area in the Weraamaia catchment indicate dynamic phases of expansion (1939–1957), inactivity (1957–1971, 1971–1984), a second phase of expansion (1984–1988), followed by inactivity (1988–1997, 1997–2003). These changes in activity are controlled by a combination of reforestation, scrub reversion, and rainfall magnitude and frequency. Gully evolution models that describe only continuous development without periods of inactivity, before ﬁnal stabilisation (e.g., Ireland et al., 1939; Sidorchuk, 1999), are not appropriate to describe gully complex development in these circumstances. Channel responses to sudden increases and decreases in sediment supply have been similarly rapid. Within the space of 30–40 years of deforestation channels had aggraded and widened prodigiously. Following progressive reforestation, and despite several large storms, including Cyclone Bola in 1988, there has been rapid incision and narrowing of channels beds. The proportion of catchment sediment yield derived from channel beds has increased from 45 to 470% in 50–60 years. The rates of sediment transfer reported in the Weraamaia and Raparapaririki streams are high compared with those reported elsewhere in the world (Knighton, 1989; Brooks and Brierley, 1997; Kondolf et al., 2002; Lie´bault et al., 2002, 2005; Owens and Walling, 2002; Surian and Rinaldi, 2004). Madej (1995) and Madej and Ozaki (1996) have reported a similar channel aggradation–degradation cycle in the 50 years following logging of the Redwood Creek catchment in northwestern California. They expect channel recovery and restoration of aquatic and riparian habitats to take several decades to a century, based on existing channel-bed elevations, pool-rifﬂe development, and sediment yields. In the Ringarooma catchment in Tasmania, 110 years of tin mining activity has also resulted in aggradation and subsequent degradation of the river. Here, upper reaches have regained their former level in about 35 years after mining ceased, while complete ﬂushing of sediment from the river is expected to take at least another 50 years (Knighton, 1989). In the Weraamaia catchment, despite sediment yields that are an order of magnitude greater, recovery based on current rates is also expected to take several decades (Kasai et al., 2005). Given the predicted effects of recent reforestation and an expected continuation of current planting rates we speculate that over the next few decades there will be a 10–20% reduction in sediment generated from hill slope sources in the Waiapu catchment. However, this reduction in the supply of sediment will result in down cutting in upper (1st–3rd order) channels. As terraces form, this will decouple hill slopes from channels. Gully fans that were supplying sediment to channels with SDRs 40.90 (Marutani et al., 1999), will become sediment storage sites. Nevertheless the vast amount of sediment stored in the channel network will ensure

Changes in basin-scale sediment supply and transfer

353

that sediment yields and impacts on the lower ﬂoodplain will remain high for many decades. Thus, although the rate of hill slope degradation will slow and in some areas show recovery, because sediment yield is regulated by storage sites that are interposed between hill slopes and the catchment outlet, there will be a lag between the change in hill slope conditions and the downstream beneﬁts (Page et al., 2000). Future trends of downstream sedimentation will be dictated by sediment supply from very large gullies that are unable to be controlled by reforestation, and sediment stored in channels, together with the rate and location of reforestation, and the magnitude and frequency of storms. Long-term patterns of sediment supply are likely to be punctuated by sudden large inputs from rare, extreme rainstorms. Clearly, targeted reforestation is the most effective management strategy to reduce both hill slope erosion and adverse downstream impacts. Effective targeting will require an improved understanding of erosion thresholds and hill slope–channel coupling. Models are needed that predict thresholds using both catchment characteristics and land-use history. This will be confounded by the fact that in dynamic landscapes thresholds and coupling relationships are changing rapidly (driven by both natural and anthropogenic factors). Predicting basin-scale responses to disturbance (sediment ﬂuxes and river behaviour) will require that local-scale coupling of landscape components must then be integrated to determine connectivity between headwater and downstream areas. Before human settlement, steep, forested headwater catchments, such as the Raparapaririki, were the major sources of sediment. During the last 100 years the major sources of sediment have shifted to deforested hills in the middle reaches of the Waiapu catchment, where gullies have provided a multitude of chronic point sources. With ongoing reforestation of these areas, the steep headwater catchments will again become the major sources of sediment, generated by spasmodic, high magnitude storm events.

Acknowledgements This paper is the result of collaboration by a number of organisations. Landcare Research was funded by the New Zealand Foundation for Research, Science and Technology (Contract C09X0013). Research in the Weraamaia catchment by Mio Kasai (Macquarie University) and Thomas Parkner (Hokkaido University) form part of PhD theses. Basil Gomez was supported by the National Science Foundation. The support of the Gisborne District Council and the Ngati Porou community, especially Tui Warmenhoven, is appreciated.

References Betts, H.D., DeRose, R.C., 1999. Digital elevation models as a tool for monitoring and measuring gully erosion. JAG (ITC) J. 1 (2), 91–101.

354

M. Page et al.

Betts, H.D., Trustrum, N.A., DeRose, R.C., 2003. Geomorphic changes in a complex gully system measured from sequential digital elevation models, and implications for management. Earth Surf. Process. Landf. 28, 1043–1058. Brooks, A.P., Brierley, G.J., 1997. Geomorphic responses of lower Bega River to catchment disturbance, 1851–1926. Geomorphology 18, 291–304. Brunsden, D., 2001. A critical assessment of the sensitivity concept in geomorphology. Catena 42, 99–123. Brunsden, D., Thornes, J.B., 1979. Landscape sensitivity and change. Trans. Inst. Br. Geogr. (New Series) 4, 463–484. Crozier, M.J., 1996. The climate-landslide couple: a Southern Hemisphere perspective. Paleocl. Res. 19 (ESF Special Issue 12), 329–350. DeRose, R.C., Gomez, B., Marden, M., Trustrum, N.A., 1998. Gully erosion in Mangatu Forest, New Zealand, estimated from digital elevation models. Earth Surf. Process. Landf. 23, 1045–1053. Fryirs, K., Brierley, G.J., 1999. Slope channel decoupling in Wolumla catchment, South Coast, New South Wales, Australia: the changing nature of sediment sources since European settlement. Catena 35, 41–63. Gomez, B., Banbury, K., Marden, M., et al., 2003. Gully erosion and sediment production: Te Weraroa Stream, New Zealand. Water Resour. Res. 39 (7), 1187, doi:10.1029/2002WR001342,2003. Harmsworth, G., Raynor, W., 2004. Cultural consideration in landslide risk perception. In: Glade, T., Anderson, M., and Crozier, M.J. (Eds), Landslide Hazard and Risk. Wiley, The Atrium, Southern Gate, Chichester, England, pp. 219–249. Harmsworth, G., Warmenhoven T., Pohatu, P., and Page, M., 2002. Waiapu Catchment Technical Report: Maori community goals for enhancing ecosystem health. Foundation for Research, Science, and Technology (FRST) contract TWWX0001. Landcare Research report LC 0102/100 for Te Whare Wananga o Ngati Porou, Ruatorea (unpublished). 185pp. Harvey, A.M., 2001. Coupling between hillslopes and channels in upland ﬂuvial systems: implications for landscape sensitivity, illustrated from the Howgill Fells, northwest England. Catena 42, 225–250. Harvey, A.M., 2002. Effective timescales of coupling within ﬂuvial systems. Geomorphology 44, 175–201. Hicks, D.L., 1995. A way to estimate the frequency of rainfall-induced mass movements (note). J. Hydrol. (NZ) 33, 59–67. Hicks, D.M., Gomez, B., Trustrum, N.A., 2000. Erosion thresholds and suspended sediment yields, Waipaoa River Basin, New Zealand. Water Resour. Res. 36 (4), 1129–1142. Hicks, D.M., Shankar, U., McKerchar, A.I., et al., 2002. River suspended sediment yields to the New Zealand coast and estuaries. Poster paper presented at 2002 New Zealand Marine Sciences Symposium, September 2002, Nelson, New Zealand. Ireland, H.A., Sharpe, C.F.S., and Eargle, D.H., 1939. Principles of gully erosion in the piedmont of South Carolina. U.S. Department of Agriculture, Washington, DC, Technical Bulletin 663, 143pp. Kasai, M., Brierley, G.J., Page, M.J., et al., 2005. Impacts of land use change on patterns of sediment ﬂux in Weraamaia catchment, New Zealand. Catena 64, 27–60, doi:10.1016/j.catena.2005.06.014. Kasai, M., Marutani, T., Reid, L.M., Trustrum, N.A., 2001. Estimation of temporally averaged sediment delivery ratio using aggradational terraces in headwater catchments of the Waipaoa River, North Island, New Zealand. Earth Surf. Process. Landf. 26, 1–16. Knighton, A.D., 1989. River adjustment to changes in sediment load: the effects of tin mining on the Ringarooma River, Tasmania, 1875–1984. Earth Surf. Process. Landf. 14, 333–359. Kondolf, G.M., Pie´gay, H., Landon, N., 2002. Channel response to increased and decreased bedload supply from land use change: contrasts between two catchments. Geomorphology 45, 35–51. Lie´bault, F., Cle´ment, P., Pie´gay, H., et al., 2002. Contemporary channel changes in the Eygues basin, southern French Prealps: the relationship of sub-basin variability to watershed characteristics. Geomorphology 45, 53–66. Lie´bault, F., Gomez, B., Page, M.J., et al., 2005. Land-use change, sediment production and channel response in upland regions. River Res. Appl. 21, 739–756, doi:10.1002/rra.880. Madej, M.A., 1995. Changes in channel-stored sediment, Redwood Creek, Northwestern California, 1947–1980. Geomorphic processes and aquatic habitat in the Redwood Creek Basin, Northwestern California. U.S. Geological Survey Professional Paper 1454, O1–O27.

Changes in basin-scale sediment supply and transfer

355

Madej, M.A., Ozaki, V., 1996. Channel response to sediment wave propagation and movement, Redwood Creek, California, USA. Earth Surf. Process. Landf. 21, 911–927. Marden, M., 2003. Waiapu gully erosion. Tairawhiti Conservation Quorum 32, 5–6. Marden, M., Arnold, G., Gomez, B., and Rowan, D., 2005. Pre- and post- reforestation gully development in Mangatu Forest, East Coast, North Island, New Zealand. River Res. Appl. 21, 757–771, doi:10.1002/rra.882. Marutani, T., Brierley, G.J., Trustrum, N.A., Page, M.J., 2001. Preface. In: Marutani, T., Brierley, G.J., Trustrum, N.A., and Page, M.J. (Eds), Source-to-Sink Sedimentary Cascades in Paciﬁc Rim GeoSystems. Matsumoto Sabo Work Ofﬁce, Ministry of Land, Infrastructure, and Transport, Motomachi, Matsumoto, Nagano, Japan, pp. 3–10. Marutani, T., Kasai, M., Reid, L.M., Trustrum, N.A., 1999. Inﬂuence of storm-related sediment storage on the sediment delivery from tributary catchments in the upper Waipaoa River, New Zealand. Earth Surf. Process. Landf. 24, 881–896. Mazengarb, C. and Speden, I.G. (compilers), 2000. Geology of the Raukumara area. Institute of Geological and Nuclear Sciences. 1:250,000 geological map 6. 1 sheet + 60 p. Lower Hutt, New Zealand. Institute of Geological and Nuclear Sciences Limited. Ministry of Forestry, 1994. A guide to the East Coast Forestry Project 1994. Ministry of Forestry, Wellington, New Zealand. Nolan, K.M. and Marron, D.C., 1995. History, causes and signiﬁcance of changes in the channel geometry of Redwood Creek, Northwestern California, 1936–1982. Geomorphic processes and aquatic habitat in the Redwood Creek Basin, Northwestern California. U.S. Geological Survey Professional Paper 1454, N1–N22. Ota, Y., Hull, A.G., Iso, N., et al., 1992. Holocene marine terraces on the northeast coast of North Island, New Zealand, and their tectonic signiﬁcance. NZ J. Geol. Geophys. 35, 273–288. Owens, P.N., Batalla, R.J., Collins, et al., 2005. Fine-grained sediment in river systems: environmental signiﬁcance and management issues. River Res. Appl. 21, 693–717, doi:10.1002/rra.878. Owens, P.N. and Walling, D.E., 2002. Changes in sediment sources and ﬂoodplain deposition rates in the catchment of the River Tweed, Scotland, over the last 100 years: the impact of climate and land use change. Earth Surf. Process. Landf. 27, 403–423, doi:10.1002/esp.327. Page, M.J., Harmsworth, G., Trustrum, N.A., et al., 2001. Waiapu River. In: Marutani, T., Brierley, G.J., Trustrum, N.A., and Page, M.J. (Eds), Source-to-Sink Sedimentary Cascades in Paciﬁc Rim Geo-Systems. Matsumoto Sabo Work Ofﬁce, Ministry of Land, Infrastructure, and Transport, Motomachi, Matsumoto, Nagano, Japan, pp. 102–111. Page, M.J., Trustrum, N.A., Gomez, B., 2000. Implications of a century of anthropogenic erosion for future land use in the Gisborne-East Coast region of New Zealand. NZ Geogra. 56 (2), 13–24. Parkner, T., Page, M.J., Marutani, T., and Trustrum, N.A., 2006. Development and controlling factors of gullies and gully complexes, East Coast, New Zealand. Earth Surf. Process. Landf. 31, 187–199, doi:10.1002/esp.1321. Patton, P.C., Schumm, S.A., 1975. Gully erosion, Northwestern Colorado: a threshold phenomenon. Geology 3, 88–90. Peacock, D.H. and Marden, M., 2004. Bed level changes in the Raparapaririki, Mangapoi, and Mangawhairiki Streams; Ruatoria. Engineering and Works Technical Report 2004/01. Gisborne District Council, Gisborne, New Zealand. Phillips, C., Marden, M., 2004. Reforestation schemes to manage regional landslide risk. In: Glade, T., Anderson, M., and Crozier, M.J. (Eds), Landslide Hazard and Risk. Wiley, The Atrium, Southern Gate, Chichester, England, pp. 517–547. Reid, L.M., Page, M.J., 2002. Magnitude and frequency of landsliding in a large New Zealand catchment. Geomorphology 49 (1–2), 71–88. Schumm, S.A., 1977. Geomorphic thresholds: the concept and its applications. Trans. Inst. Br. Geogr. (New Series) 4, 485–515. Sidorchuk, A., 1999. Dynamic and static models of gully erosion. Catena 37, 401–414. Surian, N. and Rinaldi, M., 2004. Channel adjustments in response to human alteration of sediment ﬂuxes: examples from Italian rivers. In: Sediment transfer through the ﬂuvial system, Proceedings Symposium held in Moscow, August 2004, IAHS Publication 288, 276–282.

356

M. Page et al.

Thomas, D.S.G., Allison, R.J., 1993. Landscape Sensitivity. Wiley, Chichester. Thomas, M.F., 2001. Landscape sensitivity in time and space – an introduction. Catena 42, 83–98. Trustrum, N.A., Gomez, B., Page, M.J., et al., 1999. Sediment production, storage and output: the relative role of large magnitude events in steepland catchments. Z. Geomorphol. N.F. Suppl. Bd 115, 71–86. Wainwright, J., Cases, A.C., Puigdefa´bregas, J., Michaelides, K., 2002. Linking sediment delivery from hillslope to catchment scales. Earth Surf. Process. Landf. 27, 1363–1364. Walling, D.E., Webb, B.W., 1996. Erosion and sediment yield: a global overview. IAHS Pub. 236, 3–19. Wolman, M.G., Gerson, R., 1978. Relative scales of time and effectiveness of climate in watershed geomorphology. Earth Surf. Process. 3, 189–208.

Discussion by A. Papanicolaou M. Page and his colleagues consider the geology along with the climatic and human impacts to describe changes in basin-scale sediment supply and transfer. They also focus on the response of a gully to a high magnitude rainstorm. I am disappointed that the authors have not considered other alternative methods to describe upland erosion, especially for acquiring the spatial and temporal evolution of gullies, which is a very dynamic process. The writer feels that our community should take advantage of the existing technology regarding biogeochemical tracers including radionuclides and stable isotopes. In recent years, tracing of soil erosion processes has relied heavily upon radionuclide properties of soils, including 137Cs and 7Be to study erosion rates over decades and single storms, respectively (e.g., Nagle and Ritchie, 2004). The use of biogeochemical tracers is needed in the watershed sedimentation community to better understand the source and pathway of eroded-soils within catchments and to verify watershed erosion models. It is suggested herein that stable nitrogen and carbon isotopes, d15N, d13C, and the carbon to nitrogen atomic ratio, C/N of eroded-soils may be used as natural tracers which reﬂect the erosion processes within a watershed. d15N, d13C, and C/N are biogeochemical soil properties based on carbon and nitrogen atomic arrangement within the soil. d15N is a soil property proportional to the 15N/14N isotopic ratio; d13C is proportional to the 13C/12C isotopic ratio; and C/N is ratio of total atomic carbon to total atomic nitrogen. For surface soils, which are the focus of erosion studies, carbon and nitrogen are dominated by soil organic matter (SOM) derived from decaying vegetation and plant roots. Due to this linkage, d15N, d13C, and C/N of surface soils indicate the vegetation covering the landscape and the vegetative management practices (Papanicolaou et al., 2003). Further, d15N, d13C, and C/N are dependent on soil particle size. The bivariate plot presented by Fox and Papanicolaou (2005) in a recent meeting provides an example of the versatile strength that biogeochemical tracers have in describing the source and pathways of sediments. Fig. 13.5 presents a bivariate plot of d15N vs. d13C for eroded-soil data for two events, namely, March 14–19 and March 14–April 2 (shown in ﬁgure with a triangular symbol). Mean values (7SE) for d15N and d13C samples from upland slope and ﬂoodplain locations are also included in the ﬁgure as a reference. In general the plot shows that eroded-soil from the March 14–19 was more dominated by upland rill erosion while eroded-soil from the March 14–April 2 has a greater contribution from the ﬂoodplains. The

356

M. Page et al.

Thomas, D.S.G., Allison, R.J., 1993. Landscape Sensitivity. Wiley, Chichester. Thomas, M.F., 2001. Landscape sensitivity in time and space – an introduction. Catena 42, 83–98. Trustrum, N.A., Gomez, B., Page, M.J., et al., 1999. Sediment production, storage and output: the relative role of large magnitude events in steepland catchments. Z. Geomorphol. N.F. Suppl. Bd 115, 71–86. Wainwright, J., Cases, A.C., Puigdefa´bregas, J., Michaelides, K., 2002. Linking sediment delivery from hillslope to catchment scales. Earth Surf. Process. Landf. 27, 1363–1364. Walling, D.E., Webb, B.W., 1996. Erosion and sediment yield: a global overview. IAHS Pub. 236, 3–19. Wolman, M.G., Gerson, R., 1978. Relative scales of time and effectiveness of climate in watershed geomorphology. Earth Surf. Process. 3, 189–208.

Discussion by A. Papanicolaou M. Page and his colleagues consider the geology along with the climatic and human impacts to describe changes in basin-scale sediment supply and transfer. They also focus on the response of a gully to a high magnitude rainstorm. I am disappointed that the authors have not considered other alternative methods to describe upland erosion, especially for acquiring the spatial and temporal evolution of gullies, which is a very dynamic process. The writer feels that our community should take advantage of the existing technology regarding biogeochemical tracers including radionuclides and stable isotopes. In recent years, tracing of soil erosion processes has relied heavily upon radionuclide properties of soils, including 137Cs and 7Be to study erosion rates over decades and single storms, respectively (e.g., Nagle and Ritchie, 2004). The use of biogeochemical tracers is needed in the watershed sedimentation community to better understand the source and pathway of eroded-soils within catchments and to verify watershed erosion models. It is suggested herein that stable nitrogen and carbon isotopes, d15N, d13C, and the carbon to nitrogen atomic ratio, C/N of eroded-soils may be used as natural tracers which reﬂect the erosion processes within a watershed. d15N, d13C, and C/N are biogeochemical soil properties based on carbon and nitrogen atomic arrangement within the soil. d15N is a soil property proportional to the 15N/14N isotopic ratio; d13C is proportional to the 13C/12C isotopic ratio; and C/N is ratio of total atomic carbon to total atomic nitrogen. For surface soils, which are the focus of erosion studies, carbon and nitrogen are dominated by soil organic matter (SOM) derived from decaying vegetation and plant roots. Due to this linkage, d15N, d13C, and C/N of surface soils indicate the vegetation covering the landscape and the vegetative management practices (Papanicolaou et al., 2003). Further, d15N, d13C, and C/N are dependent on soil particle size. The bivariate plot presented by Fox and Papanicolaou (2005) in a recent meeting provides an example of the versatile strength that biogeochemical tracers have in describing the source and pathways of sediments. Fig. 13.5 presents a bivariate plot of d15N vs. d13C for eroded-soil data for two events, namely, March 14–19 and March 14–April 2 (shown in ﬁgure with a triangular symbol). Mean values (7SE) for d15N and d13C samples from upland slope and ﬂoodplain locations are also included in the ﬁgure as a reference. In general the plot shows that eroded-soil from the March 14–19 was more dominated by upland rill erosion while eroded-soil from the March 14–April 2 has a greater contribution from the ﬂoodplains. The

Changes in basin-scale sediment supply and transfer

357

Figure 13.5. Mean values (7SE) for d15N and d13C samples from an upland slope and ﬂoodplain. Number in parenthesis indicates samples that were analysed in our preliminary analysis.

portend, therefore, of the tracers is that they can identify the source of the erodedsediments at the outlet of a watershed by reﬂecting the vegetative differences between ﬂoodplains and uplands and differences in sediment sizes. This method along with other methods needs to be considered in watershed monitoring and for veriﬁcation of the watershed models.

References Fox, J. and Papanicolaou, A., 2005. The Impact of agriculture erosion processes upon d15N and d13C, and C/N signatures of eroded soil. In: Parker, G. and Garcia, M.H. (Eds), River, Coastal and Estuarine Morphodynamics: RCEM 2005. Nagle, G.N., Ritchie, J.C., 2004. Wheat ﬁeld erosion rates and channel bottom sediment sources in an intensively cropped Northeastern Oregon drainage basin. Land Degradation Dev. 15, 15–26. Papanicolaou, A., Fox, J., Marshall, J., 2003. Sediment sources ﬁngerprinting in the Palouse River Watershed, USA. Int. J. Sediment Res. 18 (2), 278.

Reply by the authors This manuscript draws on studies in the Waiapu catchment, where to date biogeochemical tracers have not been used in studies of erosion. Two of the studies reported on are the work of PhD students whose methodology was developed in conjunction with their supervisors. However, in the adjacent Waipaoa catchment stable carbon isotope (d13C) and carbon to nitrogen ratio (C/N) of weathered bedrock, soil, and regolith have been

Changes in basin-scale sediment supply and transfer

357

Figure 13.5. Mean values (7SE) for d15N and d13C samples from an upland slope and ﬂoodplain. Number in parenthesis indicates samples that were analysed in our preliminary analysis.

portend, therefore, of the tracers is that they can identify the source of the erodedsediments at the outlet of a watershed by reﬂecting the vegetative differences between ﬂoodplains and uplands and differences in sediment sizes. This method along with other methods needs to be considered in watershed monitoring and for veriﬁcation of the watershed models.

References Fox, J. and Papanicolaou, A., 2005. The Impact of agriculture erosion processes upon d15N and d13C, and C/N signatures of eroded soil. In: Parker, G. and Garcia, M.H. (Eds), River, Coastal and Estuarine Morphodynamics: RCEM 2005. Nagle, G.N., Ritchie, J.C., 2004. Wheat ﬁeld erosion rates and channel bottom sediment sources in an intensively cropped Northeastern Oregon drainage basin. Land Degradation Dev. 15, 15–26. Papanicolaou, A., Fox, J., Marshall, J., 2003. Sediment sources ﬁngerprinting in the Palouse River Watershed, USA. Int. J. Sediment Res. 18 (2), 278.

Reply by the authors This manuscript draws on studies in the Waiapu catchment, where to date biogeochemical tracers have not been used in studies of erosion. Two of the studies reported on are the work of PhD students whose methodology was developed in conjunction with their supervisors. However, in the adjacent Waipaoa catchment stable carbon isotope (d13C) and carbon to nitrogen ratio (C/N) of weathered bedrock, soil, and regolith have been

358

M. Page et al.

used to ascertain the source of carbon and the extent to which POC ﬂux is tied to different erosion processes (Gomez et al., 2003). In this way most POC was found to be derived from suspended sediment generated by gully erosion (incision into weathered bedrock), and supplemented by landsliding during extreme events. Very little of the sediment generated in the Waiapu catchment is derived from surﬁcial processes (sheet wash, rill), ﬂoodplain sediments, or soil/regolith. It is overwhelmingly dominated by bedrock sources.

Reference Gomez, B., Trustrum, N.A., Hicks, D.M., et al., 2003. Production, storage, and output of particulate organic carbon: Waipaoa River Basin, New Zealand. Water Resour. Res. 39 (6), 1161, doi:10.1029/ 2002WR001619,2003.

Gravel-Bed Rivers VI: From Process Understanding to River Restoration H. Habersack, H. Pie´gay, M. Rinaldi, Editors r 2008 Elsevier B.V. All rights reserved.

359

14 Two model scenarios illustrating the effects of land use and climate change on gravel riverbeds of suburban Maryland, U.S.A. Jim Pizzuto, Glenn Moglen, Margaret Palmer and Karen Nelson

Abstract We model two scenarios consisting of 10 years of daily discharges to illustrate the effects of changing land use and climate on gravel riverbeds. The managed growth/no climate change (MGNCC) scenario represents minimum effects of land use and climate change, while the urban sprawl/climate change (USCC) scenario represents more extensive effects. We apply our scenarios to a 30 m reach of the Northwest Branch of the Anacostia River. We use downscaled precipitation estimates from the Hadley2 global circulation model (GCM) to account for climate change, and use a continuous hydrological model to produce discharge estimates. A sediment transport model, combined with empirical formulae to specify upstream sediment inputs, computes changes in grain-size distribution, bedload and suspended material discharge, suspended sediment concentration, the areal fraction of the bed in motion, bed elevation and slope, the silt–clay content of the active layer, and something new: the fraction of exposed bedrock in the active layer. The USCC scenario is characterized by larger and more frequent storm ﬂows than the MGNCC scenario, which in turn creates increased bedload and suspended load transport, increased bed mobility, and higher suspended sediment concentrations. The USCC scenario is also characterized by extreme variability in mud and bedrock content of the active layer. Convincing model predictions of the inﬂuence of climate and land use changes on gravel riverbeds will require additional study of (1) upstream sediment supply, (2) mud storage and remobilization in gravel streams, and (3) the controls on exposed bedrock in the active layer. 1.

Introduction

Changes in climate and land use both have profound effects on rivers (Palmer et al., 2002; Macklin et al., 1992; Leopold, 1968). Climate changes inﬂuence precipitation E-mail address: [email protected] (J. Pizzuto) ISSN: 0928-2025

DOI: 10.1016/S0928-2025(07)11133-0

J. Pizzuto et al.

360

and temperature, two variables that either directly or indirectly control the supply of water and sediment to stream channels. Land use changes, particularly conversion from agriculture to urban land uses, also inﬂuence storm discharges and sediment supply (Yorke and Herb, 1978). Because storms and sediment supply represent two of the most signiﬁcant controlling variables on stream morphology, land use, and climate changes are clearly important drivers in ﬂuvial geomorphology. Driven by the industrial revolution and population growth, anthropogenic inﬂuences on stream channels have probably never been greater (Hooke, 2000). These inﬂuences are primarily exerted by changes in climate and land use. For example, predictions of global climate change suggest future doubling of atmospheric CO2, increases in average temperatures by 2–61C, and increased variability in precipitation (International Panel on Climate Change (IPCC), 1997). At the same time, agricultural land in developed countries is increasingly being converted to urban and suburban land uses (e.g., Irwin and Bockstael, 2004). The profound inﬂuence of changes in climate and land use on streams has been widely noted, but methods for assessing the effects of these changes are poorly developed. In this study, we describe a preliminary effort to determine, in a speciﬁc study area, how these drivers acting together collectively inﬂuence the bed characteristics of a gravel-bed river. Our study is primarily motivated by the ultimate goal of predicting the ecological effects of changing land use and climate, so we particularly focus on variables that are important to organisms living in gravel-bed streams of our humid temperate study area (though this paper does not explicitly address the connection between hydraulic and sediment transport processes and ecological processes). This paper is designed to present our approach to modeling these processes and to present the results of modeling speciﬁc scenarios that illustrate the combined effects of both land use and climate change on gravel riverbeds. This paper does not address the relative contributions of each driver separately to changes in streambed morphology and composition. A future publication is planned to address this interesting issue.

2.

Regional setting

The Northwest Branch of the Anacostia River provides the ﬁeld setting for our study. The Northwest Branch has a drainage area of 54.6 km2, and it is located northwest of Washington, DC in Montgomery County, Maryland (detailed location maps are provided by Moore and Palmer, 2005; Palmer et al., 2002; Yorke and Herb, 1978). A United States Geological Survey (USGS) gaging station at the outlet of the basin has been active since 1924. The watershed is underlain by metamorphic rocks and is of low relief (Hunt, 1974). During the period of gaging, land use in the watershed changed from dominantly agricultural to urban. This change occurred mostly after WWII, with a particularly strong pulse of development in the late 1960s and early 1970s. During this time, peak ﬂows increased signiﬁcantly and base ﬂows decreased signiﬁcantly (Beighley and Moglen, 2002). Currently, land use in the

Two model scenarios illustrating the effects of land use and climate change

361

watershed consists of 8.6% agriculture, 24.8% forest, and 66.6% urban (Palmer et al., 2002).

3.

Methods

The goal of our research is to evaluate the inﬂuence of speciﬁc land use and climate change scenarios on stream channel bed sediment characteristics in the mid-Atlantic region. To accomplish this goal, we ﬁrst deﬁned two scenarios with differing land use and climate characteristics. One scenario we term the ‘‘Managed Growth/No Climate Change Scenario (MG/NCC)’’, while the other is referred to as the ‘‘Urban Sprawl/Climate Change Scenario (USCC)’’. These two scenarios reﬂect extreme end members of possible disturbances to ﬂuvial and stream ecological systems. The MGNCC scenario reﬂects relatively modest disturbance from land use changes, and no disturbance from climate changes. The USCC scenario reﬂects signiﬁcant disturbances from both land use and climate changes. Our approach involves comparing model predictions of ﬂuvial response to these two scenarios to help clarify the extent of disturbances from a combination of both climate and land use changes (Table 14.1). The MGNCC scenario is based on climate data from the years 1960 to 1969. Precipitation is computed using the Hadley 2 Global Circulation Model (GCM) for air temperature, with a year-round average of 17.31C. Precipitation data were estimated on a daily time step. Furthermore, because the grid spacing of the GCM is very large compared to the study area, a procedure was applied to ‘‘downscale’’ the GCM estimates to a scale more appropriate for this study (Hayhoe et al., 2004). Impervious surface is assumed to be 10%, a value below the thresholds generally associated with ecological impairments associated with urbanization, but high enough to be consistent with a signiﬁcant, densely packed urban population (Palmer et al., 2002). Forested land use in the watershed is assumed to be 20%, and the riparian zone adjacent to the stream is assumed to be forested. This level of forestation would be consistent with an urban population only with careful land use planning (Irwin and Bockstael, 2004). Finally, no construction is assumed to be taking place in the watershed. The USCC scenario is based on climate data from the years 2090 to 2099 from the Hadley 2 GCM. These predictions suggest a year-round average of 21.51C. Table 14.1.

MG/NCC US/CC

Comparison of the two climate/land-use change scenarios. Years of data

% impervious surface

% forest

Buffers

% construction

Mean air temperature (1C)

1960–1969 2090–2099

10 30

20 2

Yes No

0 2

17.3 21.5

Abbreviations: MG/NCC, managed growth/no climate change, US/CC, urban sprawl/climate change scenario.

J. Pizzuto et al.

362

Precipitation data for these years was also ‘‘downscaled’’. Impervious surface is assumed to be 30%, a value well above the thresholds associated with impaired ecological assessments of stream habitat, and consistent with the highest levels of impervious surface found at the most urban sites in the region studied by Palmer et al. (2002, 2004). Forested land use is assumed to be 1%, and forested buffers missing. This level of deforestation is seen in some of our most urban sites, although buffers are currently more intact (Palmer et al., 2002). Finally, we assume that 2% of the watershed is under construction in a given year.

3.1. 3.1.1.

Description of the sediment transport models Spatial structure and definition of selected variables

The domain of the model (Fig. 14.1) is represented by a reach of channel of length dx located within a watershed. The upstream watershed supplies a speciﬁed discharge, Q, during each time step, dt, in addition to sediment. Sediment is supplied as mud in suspension, and as sand and gravel (additional details are provided below). The channel is rectangular, and is imbedded in a ﬂoodplain of width Wf. The channel has a bankfull depth hbf. Within the reach represented by the model, the

Figure 14.1. Spatial context and geometry assumed by the model. Selected variables are also deﬁned.

Two model scenarios illustrating the effects of land use and climate change

363

ﬂoodplain width, bankfull depth, bed and water surface slope (S), and channel width (W) and rectangular cross-sectional geometry are all constant. The bed is divided into an active layer of thickness La (Parker, 1991) and a substrate that does not move. Bedrock may intrude into the active layer, and speciﬁc ‘‘rules’’ are proposed to govern how bedrock inﬂuences bedload transport processes. We have included bedrock in this way because our ﬁeld observations and those of others (Costa, 1975; Allmendinger, 2004) suggest that bedrock strongly inﬂuences bed scour and sediment transport processes in our ﬁeld area. 3.1.2.

Initial conditions

Initial values of variables obtained from ﬁeld measurements in the study are listed in Table 14.2. Several of these variables were held constant during the simulations (i.e., the active layer thickness La). Some initial values, for example, the mud content of the bed, were not measured in the ﬁeld, and these values were simply assumed to provide starting values for the computations. 3.1.3.

Boundary conditions

Boundary conditions are represented by inputs of both water and sediment. Inputs of water are determined using two models, one for precipitation and another for determination of water discharge. The supply of sediment is computed using empirical relationships based on extensive ﬁeld data. Table 14.2.

Initial values of selected variables.

Variable

Value (with comments)

Active layer thickness (m) Fraction of bedrock in active layer Fraction of pores ﬁlled with mud Bankfull depth (m) Channel roughness length Ks (m) Floodplain roughness length Kf (m) Number of grain size fractions Porosity (of mud and bed material) (%) Sediment density (kg/m3) Slope (initial) Shear stress below which mud settles (N/m2) Shear stress above which surface mud layers erode (N/m2) Spatial step – dx (m) Time step (variable, max and min given) (s) Water density (kg/m3) Width (bankfull) (m) Width of ﬂoodplain (m)

0.1 (constant throughout simulation) 0.0275 0.7 1.0 (from ﬁeld measurements) 0.2 (calibrated to measured rating curves) 0.02 (assumed) 10 30 (assumed) 2650 (assumed) 0.0089 (from ﬁeld measurements) 8.72 (assumed) 8.72 (assumed) 30 (assumed) 60–3000 (set to insure numerical stability) 1000 20.0 (from ﬁeld measurements) 200.0 (from ﬁeld measurements)

Note: Variables measured in the ﬁeld are indicated, as are variables that are held constant during the simulations.

J. Pizzuto et al.

364

The hydrological model used here, which predicts daily streamﬂow over the course of the scenarios, is completely described in McCuen and Snyder (1985). The continuous streamﬂow model used here is consistent in its conceptual structure to the Stanford Watershed Model (Crawford and Linsley, 1966), now commonly used as hydrological simulation program-Fortran (HSPF) (Bicknell et al., 1997). It supports three different forms of runoff production: surface, subsurface, and groundwater. The model requires a daily precipitation and temperature time series as well as parameter values that characterize the land use and underlying geology of the system. In separate studies (Hejazi, 2004; Palmer et al., 2002), we determined the relationships between land use and two important model parameters, namely inﬁltration versus surface runoff production, and the timing of surface runoff. Continuous streamﬂow modeling is tested by visually and quantitatively evaluating goodness-of-ﬁt between simulated and observed streamﬂow at the site of the USGS stream gages at the outlet of the NW Branch watershed. Measures of goodness of ﬁt include the correlation coefﬁcient, the Nash–Sutcliff coefﬁcient, the gain coefﬁcient, and the deviation volume. The calibration and other aspects of the hydrologic modeling are discussed in detail by Palmer et al. (2004). Correlation coefﬁcients computed for observed versus computed historical discharges, for example, ranged from 0.49 for the decade of the 1960s to 0.80 for the decade of the 1970s. Sediment inputs are speciﬁed depending on the grain size under consideration. Gravel-sized sediments are supplied that are equal to the capacity of the channel to transport each size fraction at the beginning of each time step. In practice this means that bedload transport equations are used to compute the supply of each size fraction using the sediment and hydraulic characteristics of the model reach at the beginning of each time step. This approach implies that the capacity of the channel and the supply of bed material are always in balance. While this assumption may initially appear unreasonable, it may be justiﬁed because it allows the simulations to proceed without extensive aggradation or degradation of the bed, consistent with ﬁeld observations in the area (Lewicki, 2005; Costa, 1975). Sand and silt size fractions are supplied according to empirical functions of land use and water discharge. Suspended sediment concentrations in small watersheds of the study area were extensively studied from 1964 to 1977 by Yorke and Herb (1978). They measured precipitation, changes in land use, storm discharges and concentrations in a number of watersheds in the region. Although their data represent a limited time span compared to signiﬁcant periods of climate change, they found that sediment concentrations are strongly inﬂuenced by both land use variables and storm ﬂow variables. Here we assume that relationships determined from 1964 to 1977 can be applied over different time periods, using variables that represent changes in both land use and climate. First, we obtained a correlation between the peak discharge ratio (PR) and the suspended sediment concentration divided by the percentage of construction in the watershed (raised to a power of 0.3). The peak discharge ratio is PR ¼

ðQp Qa Þ R

(14.1)

Two model scenarios illustrating the effects of land use and climate change

365

Figure 14.2. Empirical relationship between peak discharge ratio (PR) and suspended sediment concentration scaled by the mean % construction in the watershed raised to a power of 0.3. Data are illustrated for ﬁve watershed studied by Yorke and Herb (1978).

where Qp is the peak discharge during a particular storm, Qa is the discharge before the storm, and R is the total storm runoff. Yorke and Herb (1978) deﬁned Qa as the lowest discharge measured before the rising stage of an individual storm. However, in the modeling studies described below, it is represented by a constant value equal to the discharge that is equaled or exceeded by 5% of the daily discharge values. To correlate suspended sediment concentration with PR, a total of 269 observations were used. The resulting equation has an r2 of 0.39 and an F ratio that suggests a ‘‘signiﬁcance’’ well in excess of 99%. The correlation is strictly an empirical result; the exponent on the percentage of construction of 0.3, for example, was selected solely because it produced acceptable results. The empirical expression for predicting the suspended sediment concentration, C, is (Fig. 14.2) C ¼ 1:5 104 P0:3 c PR

(14.2)

where C is the concentration by weight of suspended sediment (as a ratio), and Pc the percentage of construction in the watershed during the speciﬁed time step. In the dataset used to establish equation (14.2), concentrations vary from 5 105 to 0.025, the percentage of construction varies from 0 to 15%, and the peak ratio varies from 2 to 391. In order to determine the percentages of sand and mud in suspension, empirical data presented by Yorke and Herb (1978) were used. These authors describe the variation of grain size with discharge for the Anacostia River near Colesville for 1960–1973 in their ﬁgure 11 (Yorke and Herb, 1978, p. 17). The percentage of sand increases with increasing discharge, the percentage of silt remains approximately constant at 45%, and the percentage of clay decreases with increasing discharge.

J. Pizzuto et al.

366

A simple functional representation of these data would put the % sand (Ps) at 5% for the lowest discharges, and at 30% for the highest discharges. In between the lowest and highest discharges, the percentage increases linearly on a plot of % sand as a function of log Q. These ideas were quantiﬁed using the functions summarized below. Ps ¼ 0:01 þ ð5 :01Þ PR ; PR o1 Ps ¼ 5 þ 20:762 log10 ðPR Þ; 1 PR 16 Psw ¼ 30;

PR 416

ð14:3Þ

These equations are formulated in terms of peak ratio rather than discharge in an effort to produce a non-dimensional empirical equation that could potentially be accurate when applied to watersheds of varying size. 3.1.4.

Hydraulic resistance

The relationship between water depth (H) and discharge was speciﬁed using Bray’s resistance equation for gravel-bed rivers (Chang, 1988). For ﬂows that remained in the channel, this takes the form h¼

K bs Qc pﬃﬃﬃﬃﬃﬃ aW gS

1 bþ1:5

(14.4)

where Qc is the discharge in the channel, h the water depth, g the acceleration of gravity, Ks a roughness length, and a and b coefﬁcients equal to 1.36 and 0.281, respectively. For overbank ﬂows, three equations must be must be solved by trial for h pﬃﬃﬃﬃﬃﬃ aW gS 1:5þb h (14.5) Qc ¼ K bs pﬃﬃﬃﬃﬃﬃ aW f gS ðh hbf Þ1:5þb (14.6) Qf ¼ K bsf Q ¼ Qc þ Qf

(14.7)

where Q is the total discharge, Qf is the discharge on the ﬂoodplain, and Ksf is the roughness of the ﬂoodplain. Ks was calibrated using measured stage discharge data for the study area at 0.2 m, while Ksf was arbitrarily set at 0.02 m for all computations (Table 14.2) (i.e., no attempt was made to adjust Ksf for the presence or absence of forested riparian zones). 3.1.5.

Representation of the bed

The bed in the model consists of mud, sand, gravel, and bedrock in deposits that lie below the sediment–water interface. The bed itself is divided into an active layer of thickness La in which bed material transport is localized (Parker, 1991). The active layer is considered to be ‘‘well-mixed’’ (i.e., grain-size distribution and bed porosity are treated as being uniform through the active layer). The numerical value of the

Two model scenarios illustrating the effects of land use and climate change

367

active layer thickness is presented as an input parameter in the simulation, and it does not vary with time, ﬂow conditions, or the grain size of the bed (a value of 0.1 m was used to obtain the results in this paper). The material in the active layer is divided into three different constituents: sand and gravel, bedrock, and mud. The grain-size distribution of sand and gravel is speciﬁed by denoting the fractions, fi, of speciﬁed sizes present on the bed. These fractions speciﬁcally include the fraction of bedrock exposed in the active layer. As a result, all sand and gravel grain size fractions and the fraction of bedrock must sum to 1. Mud in the bed is treated somewhat differently. The sand and gravel in the active layer are assumed to have a constant porosity. The pores in the sand and gravel are then available to be ﬁlled with mud. If the pores in the active layer become completely ﬁlled with mud, mud may be deposited on top of the active layer of sand and gravel if hydraulic conditions permit this to occur (mud transport processes are described in greater detail below). 3.1.6. Bedload transport, bed mobility, and changes in bed material grain-size distribution Bedload transport rates are computed using the equations of Wilcock and Crowe (2003). Transport rates are determined for speciﬁed size fractions from sand to gravel. In this paper, three sand size fractions and seven gravel size fractions from granule to boulders are computed. The areal extent of bedload transport, At, is an important parameter for stream ecology, and therefore the model estimates this parameter also. Methods of Wilcock (1997) are used to determine the areal extent of bed material transport at each time step. If otherwise not speciﬁed, the coefﬁcients required by these methods are taken directly from Wilcock’s (1997) laboratory studies. Changes in the bed material grain size fractions of the active layer, Fi, were computed using a modiﬁed version of the approach proposed by Parker (1991): @F i @La @q @q þ ðF i f li Þ ¼ bi þ f li bT (14.8) ð1 lp Þ La @t @t @x @x where lp is the porosity of the bed, t the time, x the downstream spatial coordinate, qbi is the volumetric bed material transport rate of grain size fraction i, qbt is the total volumetric bed material transport rate (i.e., qbi summed over all grain sizes), and fli is the fraction of grain size i at the interface between the active layer and the substrate. The value of fli is set equal to 0 if the fraction of the bedrock (Fbrf) in the active layer is positive. If Fbrf ¼ 0, then methods of Parker are used to determine fli. During periods of bed aggradation and degradation, sediment transfers between the load, substrate, and the active layer are computed using the weighted average approach described by Cui et al. (1996). 3.1.7.

Changes in the fraction of bedrock in the active layer

If equation (14.8) is rederived with bedrock as a constituent of the active layer, and if the resulting equation is summed over all grain sizes, then equation (14.9) is

J. Pizzuto et al.

368 obtained: ð1 lp Þ

@La @F brf @Z @q ðf brfl F brf Þ ð1 lp ÞLa þ ð1 lp Þð1 f brfl Þ ¼ bT @t @t @x @t (14.9)

where fbrﬂ is the bedrock fraction at the interface between the substrate and the active layer. According to equation (14.9), the presence of bedrock in the active layer results in an interesting set of possibilities when gradients in total bedload transport exist. For example, an excess of supply over transport capacity could result in sediment storage in the reach. Equation (14.9) suggests that this storage could be accommodated in one of the three possible ways: (1) deposition on the bed (i.e., positive (qZ/qt), (2) a decrease in bedrock fraction in the active layer (i.e., negative qFbrf/qt), or (3) changes in the thickness of the active layer, modulated by a difference between the bedrock fraction in the active layer and that at the boundary of the substrate (ﬁrst term on the left of equation (14.9)) (this third option, of course, was not explicitly considered during this study because La is treated as a constant during the present simulations). It is important to note that current understanding of partially alluvial channels is not sufﬁcient to predict which of these three outcomes will occur, or if some combination of the three is likely. In the absence of a clear understanding of how to ‘‘partition’’ sediment transport gradients between the three terms on the left of equation (14.9), several ‘‘rules’’ are adopted here. First, it is assumed that the thickness of the active layer, La, remains constant during the simulation. This removes the ﬁrst term on the left of equation (14.9). Second, it is assumed that when Fbrf40, imbalances in sediment ﬂux are accommodated by changes in bedrock fraction (i.e., the middle term on the left of equation (14.9)) rather than by changes in the elevation of the bed. It is only when the channel becomes fully alluvial, i.e., when Fbrf ¼ 0, that gradients in bedload transport are accommodated by changes in bed elevation. The arbitrary nature of these rules should be noted. Further studies, presumably in the laboratory, are urgently needed to better deﬁne how gradients in sediment transport lead to changes in bed elevation and the fraction of bedrock in the active layer. Equation (14.9) is not actually implemented in the numerical model – it is presented here to clarify the rules used to determine changes in Fbrf. Fbrf is actually computed by summing all the other grain size fractions, Fi, during a time step (after these have been determined by solving equation (14.8), and noting that all of the grain size fractions AND the bedrock fraction must sum to 1. 3.1.8.

Computing washload concentrations

The washload concentration Cm is determined by solving a mass balance equation of suspended ﬁne-grained sediment. This equation is: @ðhC m Þ 1 @ðQC m Þ ¼ DðtÞ þ EðtÞ @t W @x

(14.10)

Two model scenarios illustrating the effects of land use and climate change

369

where W is the channel width and D and E rates of deposition and erosion of mud, respectively. Both D and E have units of volume of mud eroded per unit bed area per unit time (m3/m2/s). 3.1.9.

Specifying rates of erosion and deposition of mud on the bed

Storage of ‘‘ﬁne’’ sediment in gravel riverbeds has been investigated in the ﬁeld and laboratory (Rehg et al., 2005; Packman and McKay, 2003; Carling, 1984; Diplas and Parker, 1992; Jobson and Carey, 1989; Lisle, 1989; Reid and Frostick, 1985; Frostick et al., 1984, and many others). These studies have indicated that a variety of variables control the extent of storage and remobilization of ﬁne-grained sediment, including the grain-size distributions of the bed and the ‘‘ﬁnes’’, the concentration of ﬁnegrained sediment in transport, the chemical properties of the ﬁne-grained sediment, the extent and intensity of bed material transport, the nature of hyporheic exchange between the ﬂow and the bed, and the bed topography viewed at a variety of spatial scales. In some cases, deposition of ﬁne-grained sediment is concentrated in a thin surface layer of the bed, while in other cases, deposition occurs to considerable depth within the bed material. Surface layers of ﬁne-grained sediment may be observed on the bed under certain conditions, while deposition is limited to the interstitial openings under other conditions. The deﬁnition of ‘‘ﬁne’’ sediment is particularly broad, and often includes sand sizes in addition to silt and clay. General, quantitative models to evaluate storage and remobilization of mud in gravel beds are lacking. Because the processes of ﬁne-grained sediment erosion and deposition over gravel beds are poorly understood, and to produce a relatively simple model that does not require a large number of unknown parameters, a series of ‘‘rules’’ are followed here to represent these processes. Rates of deposition and erosion are speciﬁed as simple functions of the bed shear stress, t: D ¼ C m V s ðtÞ

(14.11)

E ¼ EðtÞ

(14.12)

where Vs(t) is the settling velocity of mud. Vs is speciﬁed as V s ¼ 0; V s ¼ V ss

t4tcd t tcd

(14.13) (14.14)

here tcd is a critical boundary shear stress for mud deposition (Partheniades, 1986) and Vss the settling velocity of mud-sized sediment under conditions of deposition. Deposited sediment can be stored in two places in the bed. If the pores of the active layer are not completely ﬁlled with mud, then the volume of mud deposited during a time step can be placed in the pores. If the pores are completely ﬁlled (either by previous deposition, or by part of the mud deposited during a current time step), then mud is deposited on top of the sand and gravel active layer as a layer of pure mud (i.e., mud without any sand and gravel within it). The erosion rate E(t) is given by E¼

V me Dt

(14.15)

J. Pizzuto et al.

370

where Vme is the volume of mud eroded per unit bed area during a particular time step. The volume of mud eroded is determined by a series of rules. If a layer of mud is present, and if the boundary shear stress exceeds a critical threshold for mud erosion tce, then the entire mud layer is removed, and this volume contributes to Vme. If gravel is in motion, mud can also be removed from the pores of the active layer. Mud is removed from pores when the fraction of pores ﬁlled with mud is greater than the total fraction of the active layer that is NOT immobile. This implies that the portion of the gravel layer without mud in its pores is moved ﬁrst. If, however, some fraction of the active layer is in motion that contains mud, all the mud in these pores is moved during a time step. These rules lead to the following quantitative expressions for the solid volume of mud removed from the bed during a time step: V me ¼ V ms þ V mp

(14.16)

where Vms is the volume of the surface layer of mud removed (if any) (per unit bed area) and Vmp the volume of mud per unit bed area eroded from pores. Vms and Vmp are speciﬁed by V ms ¼ T s ð1 lm Þ

(14.17)

V mp ¼ f c La ð1 lm Þ

(14.18)

where Ts is the thickness of the mud layer, lm the porosity of the mud deposit, and fc, the fraction of the active layer ‘‘cleaned’’ by erosion, is given by fc ¼ 0

At f pm

f c ¼ At f pm ;

At 4f pm

(14.19) (14.20)

where fpm is the fraction of the pores ﬁlled with mud at the beginning of the current time step. 3.1.10.

Numerical methods and programming

The two differential equations presented above (equations (8) and (10)) are discretized using simple explicit ﬁnite different approximations. The program is written as a MATLAB script and executed in MATLAB version 5.2. A spatial step of 30 m was used in all computations. The time step was varied to insure accuracy and numerical stability. For the lowest ﬂows, a time step of 3000 s proved adequate. For higher ﬂows, the time step was decreased stepwise with increasing discharge in four steps to a minimum value of 60 s. 3.2.

Sensitivity analysis

Because many of the parameters required by the model are poorly constrained by either ﬁeld observations or even scientiﬁc understanding, we performed a sensitivity analysis to document how our results might be inﬂuenced by uncertainty in estimating input variables and parameters. The sensitivity analysis was based on a 2-year

Two model scenarios illustrating the effects of land use and climate change

371

simulation of daily ﬂows using the USCC scenario. Independent variables selected for sensitivity analysis included input values of the daily discharge Q, the thickness of the active layer La, the roughness length of the channel bed Ks, the bed porosity lp, the critical shear stress for mud deposition (and erosion) tcd, the concentration of suspended sediment supplied by from the watershed C, and the volume of bed material supplied by the watershed for each grain size fraction qbi. During the sensitivity analysis, each of these independent variables was increased by 50% from ‘‘base’’ values used in the USCC simulation (the volume of bed material, however, was decreased by 50% because of numerical issues that arose when the model was run using increased bedload transport rates). For variables such as discharge that change with each day of the simulation, all the daily values were increased (or decreased) by 50%. For each independent variable, model sensitivity was assessed by computing mean and 90th percentiles for the 730 daily values of four dependent variables: the fraction of exposed bedrock in the active layer (Fbrf), the areal fraction of the bed mobilized, the fraction of mud ﬁlled with pores (fpm), and the concentration of suspended wash load Cm. Differences between the ‘‘base’’ simulation and the simulations obtained by varying individual parameters were expressed as % differences according to the following formula: % difference ¼

100 ðstatistic from sensitivity simulation statistic from base simulationÞ ðstatistic from base simulationÞ

(14.21) where the ‘‘statistic’’ in equation (14.21) could refer to either the median or the 90th percentile value for the parameter in question.

4.

Results

The total precipitation varied considerably between the MGNCC and the USCC scenarios. For the MGNCC scenario, a total of 2601 mm occurred in the 10-year period between 1960 and 1969. For the USCC scenario, a total of 3193 mm occurred between 2090 and 2099. The greater precipitation estimated for the USCC scenario is reﬂected in ﬂow duration curves of daily discharges obtained from hydrological modeling (Fig. 14.3). Only the upper ends of the ﬂow duration curves are illustrated in Fig. 14.3, because at lower durations, the two curves are indistinguishable. This indicates that the USCC scenario is characterized by higher storm ﬂows than the MGNCC scenario. It is useful to note that discharge in Fig. 14.3 is illustrated on a logarithmic axis, so the storm ﬂows for the USCC scenario are substantially larger than those of the MGNCC scenario. The increased storm discharges of the USCC are not surprisingly reﬂected in increased volumes of bedload transported out of the study reach during the 10-year simulation period. The results of the simulations (not presented here due to space limitations) indicate that the total volume of bed material transported under the USCC scenario is more than two times the volume transported under the MGNCC

372

J. Pizzuto et al.

Figure 14.3. Flow duration curve comparing the frequency of discharges for the ‘‘Urban Sprawl with Climate Change’’ scenario with those of the ‘‘Managed Growth/No Climate Change’’ scenario. Flow duration curves are truncated at a frequency of 0.95. At lower frequencies, differences between the curves are indistinguishable.

scenario. Furthermore, most of the sediment is transported during six or seven large storms. ‘‘Duration’’ curves illustrating the areal fraction of the bed that is mobile during transport events for the two scenarios are illustrated in Fig. 14.4. Not surprisingly, the larger ﬂoods of the USCC scenario are associated with greater areal bed mobility than the MGNCC scenario. At a duration of 50%, for example, the USCC scenario predicts 65% areal bed mobility, while the MGNCC scenario is associated with 62% areal bed mobility. The maximum bed mobility for the USCC scenario is 83%, while that for the MGNCC scenario is 76%. Box plots inset into Fig. 14.4 illustrate the distributions of bed mobility for the MGNCC and USCC scenarios (outliers, however, are omitted). A non-parametric Kruskal–Wallis analysis yields a p value of less than 0.001, suggesting that the median values of these distributions are signiﬁcantly different. The dynamics of the bedrock fraction of the active layer are very different for the two scenarios (Fig. 14.5). For the MGNCC scenario, the exposed bedrock fraction gradually increases during the 10-year simulation from 2.8 to 8%, reﬂecting a reduced sediment transport supply relative to the sediment transport capacity of the stream that is likely related to the current condition of this urbanized watershed. For the USCC, the exposed bedrock fraction varies rapidly between 0 and 12%, apparently reﬂecting extremes of supply and transport capacity during the largest ﬂoods. However, for most of the time, the bedrock fraction of the active layer during the USCC simulation is lower than that of the MGNCC simulation. Initial and ﬁnal grain-size frequency curves for the two scenarios are surprisingly similar for the two scenarios (though the details are not presented here). Results are

Two model scenarios illustrating the effects of land use and climate change

373

Figure 14.4. Duration curves for areal fraction of bed mobility for the two scenarios. The inset shows box plots for the distributions of areal bed mobility fractions for the two scenarios.

Figure 14.5. Time series of the fraction of exposed bedrock for the two scenarios.

374

J. Pizzuto et al.

Figure 14.6. Duration curves for volumetric suspended sediment concentration for the two scenarios. The inset shows box plots for the distributions of sediment concentrations for the two scenarios.

obtained for ﬁne sand up to boulders 500 mm in diameter. Grain-size frequency curves for both scenarios suggest considerable changes in bed material grain-size distribution during the 10-year simulations. These changes involve a reduction in the fraction of sand sizes on the bed, as well as a reduction in the fractions of the largest particles on the bed. Remarkably, both scenarios present very similar grain-size distributions at the ends of the simulations. Not surprisingly suspended sediment concentrations for the USCC scenario are signiﬁcantly higher than those of the USCC scenario, reﬂecting the greater sediment supply associated with storms in the USCC scenario (Fig. 14.6). A non-parametric Kruskal–Wallis analysis of these distributions yields a p value of less than 0.001, suggesting that the median values of the suspended sediment concentrations are drawn from signiﬁcantly different populations. The fractions of the active layer ﬁlled with mud are compared for the two scenarios in Fig. 14.7. This parameter hovers around 40% for the MGNCC scenario, with occasional excursions from 65% to lows of 20–30%. For the USCC scenario, much greater variability is encountered, with high values of 100% occurring approximately seven times, and a series of low values of less than 20% occurring a similar number of times. Average values for the USCC scenario appear similar to those of the MGNCC scenario. Results of sensitivity analyses are presented in Fig. 14.8, where percent differences between the base simulation and simulations with 50% changes to seven parameters are tabulated. Shading indicates four categories of sensitivity, from very low (white)

Two model scenarios illustrating the effects of land use and climate change

375

Figure 14.7. Time series of the fraction of pores ﬁlled with mud for the two scenarios.

to very high (black). The model appears to be very insensitive to uncertainty in the thickness of the active layer, the roughness length, and the bed porosity. The model is moderately sensitive to changes in discharge. For example, the fraction of bedrock increases when all the discharges are increased by 50% due to scour of the bed, and the 90th percentile fraction of pores ﬁlled with mud also decreases, presumably due to the same process. However, bed mobility and wash load concentration are insensitive to changes in discharge. Two independent variables, tcd, and the concentration of suspended sediment supplied from the watershed represent variables that speciﬁcally represent ﬁnegrained sediment dynamics in the model. Not surprisingly, variations in these parameters do not inﬂuence the bedrock fraction of the active layer or bed mobility. The fraction of pores ﬁlled with mud and the concentration of wash load are, however, moderately inﬂuenced by changes in these parameters, particularly in the extremes of the distributions represented by the 90th percentile statistics. Remarkably, the model results prove to be most sensitive to the volume of bed material supplied from the watershed. All of the dependent variables show considerable sensitivity to bed material supply. This sensitivity to the imposed decrease of 50% arises because the lack of sediment at all discharges results in scour of the bed, which (1) increases the fraction of exposed bedrock in the active layer, (2) coarsens the bed, thereby reducing bed mobility, (3) replaces unﬁlled pores with bedrock, increasing the fraction of pores ﬁlled with mud, and (4) increases wash load concentrations by scouring mud-ﬁlled porous sediment from the bed.

J. Pizzuto et al.

376

Figure 14.8. Results of sensitivity analyses. Percent differences from the base simulation are shown for the median and also for the 90th percentile. The sensitivity of four dependent variables to changes in seven independent variables are assessed. Results are shaded according to four categories of sensitivity.

5. 5.1.

Discussion Interpretations of the results

The results presented above can be divided into two categories. One set of results may be derived directly from the higher discharges associated with the USCC

Two model scenarios illustrating the effects of land use and climate change

377

scenario. Not surprisingly, these include increased bedload transport, increased bed mobility, and increased suspended sediment concentrations. All of these results are easily associated with higher discharges. The other set of results includes the increased variability in mud and bedrock fractions in the active layer. These results are caused by variations in the balance between sediment supply and transport capacity, although the detailed mechanisms involved differ somewhat. The increased variability in mud content of the active layer arises from two related phenomena. First, storm discharges are associated with signiﬁcantly higher sediment concentrations in the USCC scenario as compared to the MGNCC scenario due to the higher percentage of construction in the USCC scenario (i.e., Equation (14.2)). Flows smaller than the largest stormﬂows (e.g., those in Fig. 14.3 with a duration of 0.97) occur much more frequently in the USCC scenario, and they carry more sediment than similar duration ﬂows of the MGNCC scenario. At these ﬂows, however, hardly any of the bed is in motion, and therefore extensive amounts of mud are free to accumulate in the pores. Then, when the highest storm ﬂows do occur, nearly all of this mud is removed as most of the active layer is mobilized. This leads to very high levels of variability in mud content of the active layer in the USCC scenario. In the MGNCC scenario, mud accumulates to a lesser degree because of the lower concentrations (equation (14.2)), and the extent of erosion is also less because of the reduced mobility of the bed (Fig. 14.4). Both of these processes lead to lower variability in mud content in the MGNCC scenario. Variability in bedrock fraction arises from changes in the balance between upstream sediment supply and channel transport capacity during the hydrographs of individual storms. First, it is important to realize that the continuous hydrological model predicts that storms last for several days. Thus, each ‘‘high ﬂow event’’ actually consists of a complete hydrograph with rising and falling limbs that involves a number of daily ﬂows. When model predictions are carefully analyzed for individual storm hydrographs, the following processes are revealed. During the initial rapid rise of ﬂow to the peak discharge of the storm, sediment transport capacity is greater than upstream sediment supply, and as a result, sediment is removed from the active layer, and the fraction of bedrock in the active layer rapidly increases. However, during the falling limb of the storm, the relationship between capacity and supply is reversed, such that supply exceeds capacity, causing sediment to accumulate and the bedrock fraction to gradually decrease. Because both sediment supply and capacity increase signiﬁcantly during high ﬂows, these imbalances between supply and capacity cause large changes in bedrock exposure during the largest storms. As a result, the larger storm ﬂows of the USCC scenario cause much larger variability in bedrock fraction of the active layer than the somewhat smaller storm ﬂows of the MGNCC scenario.

5.2. Implications for modeling geomorphic and ecological consequences of climate and land use changes The results presented above highlight important areas where additional research is needed to provide an improved capability for predicting the responses of riverbeds to

378

J. Pizzuto et al.

climate changes. These include (1) better methods to predict upstream sediment supply, (2) improved understanding of the partitioning between deposition, erosion, and bedrock exposure of the active layer, (3) better models of wash load storage in river channels, and (4) extension of the model to include channel widening and ﬂoodplain accumulation. It is signiﬁcant that differences in variability in mud and bedrock fractions in the active layer both arise from the interactions between sediment supply from upstream and transport capacity of the channel. It is also signiﬁcant that this variability is very important for ecological processes (particularly the mud content of the active layer) (e.g., Wood and Armitrage, 1997). However, this variability arises from subtle relationships between supply and transport capacity that are rather poorly constrained. The equations for upstream supply of suspended sediment have extremely large amounts of scatter themselves, as is typical of suspended sediment data. Upstream supply of gravel is essentially arbitrarily determined by the capacity of the ﬂow, a ‘‘rule’’ that was adopted in this study primarily because a more convincing approach is unavailable. Because the substantial variability in mud and bedrock content is caused by subtle variations in supply and transport capacity, and because supply in particular is poorly known, one wonders if the model results simply reﬂect an arbitrarily and poorly constrained knowledge of sediment supply, rather than a well-constrained prediction. Further study of sediment supply processes is clearly important and should be a high priority, particularly given the sensitivity of the computations to assumptions and parameter values that control the supply of suspended sediment and bed material (Fig. 14.8). The computation of the fraction of bedrock in the active layer is also based on some rather arbitrary rules. These rules are used to determine whether deposited sediment is used to increase the elevation of the bed or to cover exposed bedrock. In the present study, deposition covers bedrock until the bedrock content of the active layer falls to zero. When bedrock is no longer present in the active layer, then increases in bed elevation are allowed. These rules, however, are essentially arbitrary, and other ‘‘rules’’ are possible and even likely. A careful ﬂume study of these processes would be particularly useful in guiding further development of models of ‘‘partially alluvial’’ rivers. Storage of mud in the pores of the bed is a particularly important ecological process. However, this process has not been extensively studied, and as a result a simple approach has been adopted in this paper. However, a more convincing method would be desirable, particularly given the sensitivity of mud pore storage processes in the model to sediment supply parameters that are themselves poorly constrained and understood. The model used in this paper essentially only includes processes that occur on the bed of the stream. However, ﬁeld observations indicate that the decadal sediment budget of these stream channels is inﬂuenced by ﬂoodplain accumulation and channel widening (Allmendinger, 2004). A more complete model would also include these processes.

Two model scenarios illustrating the effects of land use and climate change 5.3.

379

Assessment of model results

Given the limitations of the present models discussed above, how should the model results presented in this paper be viewed? Are some of the results robust and convincing, and if so, which ones? The predictions related to increased storm ﬂows are probably convincing results of climate change and urban sprawl. These include increased suspended sediment concentrations, increased bed mobility, and increased transport of bed material downstream. However, the predictions presented here should be viewed as qualitative trends, rather than as absolute numerical predictions, because the models used here have not been directly calibrated for ﬁeld conditions. Other predictions should be evaluated with some caution. The predicted grain-size composition of the bed appears to be relatively insensitive to land use and climate changes; however, these computations require accurate prediction of transport of individual size fractions ranging from sand to boulders, and these computations are typically not very precise. Predictions that result from the balance between sediment transport capacity and upstream supply (e.g., mud and bedrock content of the active layer) are also subject to high uncertainty, and the sensitivity analysis clearly indicates that many model predictions are very sensitive to variations in the supply of wash load and bed material. Thus, many of these results should be viewed as hypotheses that require additional testing and model development before being accepted with conﬁdence.

6.

Conclusions

In this paper we have reported model results for two scenarios of combined land use and climate change. The models are designed to predict changes in bed elevation, slope, grain-size distribution, bedrock exposure, and sediment transport processes for a study area of the Northwest Branch of the Anacostia River in Montgomery County, Maryland, just northwest of Washington, DC. The models provide input for predicting the effects of land use and climate change on stream ecology, so particular attention was given to variables of interest to stream ecologists. The two scenarios reﬂect end members of (1) minimum disturbance due to land use and climate change (the MGNCC scenario) and (2) maximum disturbance (the USCC scenario). The model results indicate that the USCC scenario is associated with larger, more frequent storm discharges than the MGNCC scenario. The increased storm discharges of the USCC scenario create increased bed material transport and bed disturbance, and higher suspended sediment concentrations than those of the MGNCC scenario. Furthermore, the USCC scenario is also associated with highly variable mud and bedrock contents in the active layer. This variability is created by changes in the balance between sediment supply and transport capacity that

380

J. Pizzuto et al.

occur during individual storm hydrographs. The mud content of the active layer is particularly important for a variety of ecological processes. Several important processes represented by our models should be studied in greater detail. These include (1) processes that deposit and erode mud from the active layer of the bed, (2) the inﬂuence of bedrock on sediment transport and storage in a reach, and (3) processes that supply sediment of all grain sizes to a reach. Future models that represent similar processes on decadal timescales should also include changes in channel width and overbank sedimentation.

Acknowledgments Funding and support for this project was provided by the U.S. Environmental Protection Agency, including the Science to Achieve Results (STAR) Program (EPA number: R828012) and the Global Climate Change Program (EPA numbers: 1W0594NAEX and R83038701). Two anonymous reviewers provided very helpful comments.

References Allmendinger, N., 2004. The inﬂuence of convex-bank ﬂoodplains on stream channel processes and morphology in the Mid-Atlantic Piedmont. Ph.D. Thesis, University of Delaware, Newark, DE, 223pp. Beighley, R.E., Moglen, G.E., 2002. Assessment of stationarity in rainfall-runoff behavior in urbanizing watersheds. J. Hydrolog. Eng. ASCE 7, 27–34. Bicknell, B.R., Imhoff, J.C., Kittle, J.L., et al., 1997. Hydrologic simulation program – Fortran. User’s manual for version 11. EPA/600/R-97/080, E.S. EPA, Ecosystems Research Division, Environmental Research Laboratory, Athens, GA. Carling, P.A., 1984. Deposition of ﬁne and coarse sand in an open-work gravel bed. Can. J. Fish. Aquat. Sci. 41, 263–280. Chang, H., 1988. Fluvial processes in river engineering. Wiley, New York, 432pp. Costa, J., 1975. Effects of agriculture on erosion and sedimentation in the Piedmont Province, Maryland. Geol. Soc. Am. Bull. 86, 1281–1286. Crawford, N.H., Linsley, R.K., 1966. Digital simulation in hydrology: Stanford Watershed Model IV. Technical Report 39, Department of Civil Engineering, Stanford University, Stanford, CA. Cui, Y., Parker, G., Paola, C., 1996. Numerical simulation of aggradation and downstream ﬁning. J. Hydraul. Res. 34, 185–204. Diplas, P., Parker, G., 1992. Deposition and removal of ﬁnes in gravel-bed streams. In: Billi, P., Hey, R.D., Thorne, C.R., and Tacconi, P. (Eds), Dynamics of Gravel-Bed Rivers. Wiley, London, pp. 313–329. Frostick, L.E., Lucas, P.M., Reid, I., 1984. The inﬁltration of ﬁne matrices into coarse-grained alluvial sediments and its implications for stratigraphical interpretations. J. Geol. Soc. London 141, 955–965. Hayhoe, K., Cayan, D., Field, C.B., et al., 2004. Emission pathways, climate change, and impacts on California. Proc. Natl. Acad. Sci. USA 101, 12422–12427. Hejazi, M., 2004. The joint effects of climate change and urbanization on the distribution of streamﬂow magnitudes in the Maryland piedmont region. University of Maryland M.S. thesis, 179pp. The complete citation and, indeed, the manuscript itself, may be obtained from, http://hdl.handle.net/1903/ 1547. Hooke, R.L., 2000. On the history of humans as geomorphic agents. Geology 28, 843–846. Hunt, C.B., 1974. Natural regions of the United States and Canada. Freeman, San Francisco, CA, 725pp.

Two model scenarios illustrating the effects of land use and climate change

381

International Panel on Climate Change (IPCC), 1997. The regional impacts of climate change: An assessment of vulnerability. In: Watson, R.T., Zinyowera, M.C., and Moss, R.H. (Eds), Report of the IPCC Working Group II. Cambridge University Press, Cambridge, p. 517. Irwin, E., Bockstael, N., 2004. Land use externalities, open space preservation, and urban sprawl. Reg. Sci. Urban Econ. 34, 705–725. Jobson, H.E., Carey, W.P., 1989. Interaction of ﬁne sediment with alluvial streambeds. Water Resour. Res. 25, 135–140. Leopold, L.B., 1968. Hydrology for urban land planning: A guidebook on the hydrologic effects of urban land use. US Geol. Surv. Circular 554, 18. Lewicki, M., 2005. A watershed scale numerical model of the impact of land use change on bed material transport in suburban Maryland. Ph.D. Thesis, University of Delaware, Newark, DE, 274pp. Lisle, T.E., 1989. Sediment transport and resulting deposition in spawning gravels, north coastal California. Water Resour. Res. 26, 1303–1319. Macklin, M.G., Rumsby, B.T., Heap, T., 1992. Flood alluviation and intrenchment: Holocene valley ﬂoor development and transformation in the British uplands. Geol. Soc. Am. Bull. 104, 631–643. McCuen, R.H., Snyder, W.M., 1985. Hydrologic modeling: Statistical methods and applications. PrenticeHall, Englewood Cliffs, NJ. Moore, A.M., Palmer, M.A., 2005. Invertebrate biodiversity in agricultural and urban headwater streams: implications for conservation and management. Ecol. Appl. 15 (4), 1169–1177. Packman, A.I., McKay, J.S., 2003. Interplay of stream-subsurface exchange, clay particle deposition, and streambed evolution. Water Resour. Res. 39(4), 1097, doi:10.1029/2002WR001432. Palmer, M.A., Bockstael, N., Moglen, G., et al., 2004. Spatial patterning of land use conversion: Linking economics, hydrology, and ecology to evaluate the effects on stream ecosystems. Final technical report, U.S. EPA Agreement No. R-82801021, 17pp. Palmer, M.A., Moglen, G.E., Bockstael, N.E., et al., 2002. The ecological consequences of changing land use for running waters: The suburban Maryland case. Yale Bull. Environ. Sci. 107, 85–113. Parker, G., 1991. Selective sorting and abrasion of river gravel. 1. Theory. J. Hydraul. Eng. 117(2), 131–149. Partheniades, E., 1986. A fundamental framework for cohesive sediment dynamics. In: Mehta, A.J. (Ed.), Estuarine Cohesive Sediment Dynamics, Lecture Notes on Coastal and Estuarine Studies. SpringVerlag, Berlin, Vol. 14, pp. 219–250. Rehg, K.J., Packman, A.I., Ren, J., 2005. Effects of suspended sediment characteristics and bed sediment transport on streambed clogging. Hydrolog. Process. 19, 413–427. Reid, I., Frostick, L.E., 1985. Role of settling, entrainment and dispersive equivalence and of interstice trapping in placer formation. J. Geol. Soc. London 142, 739–746. Wilcock, P.R., 1997. The components of fractional transport rate. Water Resour. Res. 33, 247–258. Wilcock, P.R., Crowe, J.C., 2003. Surface-based transport model for mixed-size sediment. J. Hydraul. Eng. 129, 120–128. Wood, P.J., Armitrage, P.D., 1997. Biological effects of ﬁne sediment in the lotic environment. Environ. Manage. 21, 203–213. Yorke, T.H., Herb, W.J., 1978. Effects of urbanization on streamﬂow and sediment transport in the Rock Creek and Anacostia River Basins, Montgomery County, Maryland, 1962–1974. US Geol. Surv. Prof. Pap. 1003, 71.

Discussion by Bob Mussetter An important assumption in the model is that the mud must ﬁll pore spaces in the bed-material matrix before it will deposit on the bed surface. This assumption has a signiﬁcant effect on predicted mud dynamics. Our experience in the Upper Colorado River system is that mud deposits only on the surface and does not inﬁltrate in signiﬁcant quantities into the bed. As a result, the mud is much more transient than would be predicted by this model. There may be differences in physical

Two model scenarios illustrating the effects of land use and climate change

381

International Panel on Climate Change (IPCC), 1997. The regional impacts of climate change: An assessment of vulnerability. In: Watson, R.T., Zinyowera, M.C., and Moss, R.H. (Eds), Report of the IPCC Working Group II. Cambridge University Press, Cambridge, p. 517. Irwin, E., Bockstael, N., 2004. Land use externalities, open space preservation, and urban sprawl. Reg. Sci. Urban Econ. 34, 705–725. Jobson, H.E., Carey, W.P., 1989. Interaction of ﬁne sediment with alluvial streambeds. Water Resour. Res. 25, 135–140. Leopold, L.B., 1968. Hydrology for urban land planning: A guidebook on the hydrologic effects of urban land use. US Geol. Surv. Circular 554, 18. Lewicki, M., 2005. A watershed scale numerical model of the impact of land use change on bed material transport in suburban Maryland. Ph.D. Thesis, University of Delaware, Newark, DE, 274pp. Lisle, T.E., 1989. Sediment transport and resulting deposition in spawning gravels, north coastal California. Water Resour. Res. 26, 1303–1319. Macklin, M.G., Rumsby, B.T., Heap, T., 1992. Flood alluviation and intrenchment: Holocene valley ﬂoor development and transformation in the British uplands. Geol. Soc. Am. Bull. 104, 631–643. McCuen, R.H., Snyder, W.M., 1985. Hydrologic modeling: Statistical methods and applications. PrenticeHall, Englewood Cliffs, NJ. Moore, A.M., Palmer, M.A., 2005. Invertebrate biodiversity in agricultural and urban headwater streams: implications for conservation and management. Ecol. Appl. 15 (4), 1169–1177. Packman, A.I., McKay, J.S., 2003. Interplay of stream-subsurface exchange, clay particle deposition, and streambed evolution. Water Resour. Res. 39(4), 1097, doi:10.1029/2002WR001432. Palmer, M.A., Bockstael, N., Moglen, G., et al., 2004. Spatial patterning of land use conversion: Linking economics, hydrology, and ecology to evaluate the effects on stream ecosystems. Final technical report, U.S. EPA Agreement No. R-82801021, 17pp. Palmer, M.A., Moglen, G.E., Bockstael, N.E., et al., 2002. The ecological consequences of changing land use for running waters: The suburban Maryland case. Yale Bull. Environ. Sci. 107, 85–113. Parker, G., 1991. Selective sorting and abrasion of river gravel. 1. Theory. J. Hydraul. Eng. 117(2), 131–149. Partheniades, E., 1986. A fundamental framework for cohesive sediment dynamics. In: Mehta, A.J. (Ed.), Estuarine Cohesive Sediment Dynamics, Lecture Notes on Coastal and Estuarine Studies. SpringVerlag, Berlin, Vol. 14, pp. 219–250. Rehg, K.J., Packman, A.I., Ren, J., 2005. Effects of suspended sediment characteristics and bed sediment transport on streambed clogging. Hydrolog. Process. 19, 413–427. Reid, I., Frostick, L.E., 1985. Role of settling, entrainment and dispersive equivalence and of interstice trapping in placer formation. J. Geol. Soc. London 142, 739–746. Wilcock, P.R., 1997. The components of fractional transport rate. Water Resour. Res. 33, 247–258. Wilcock, P.R., Crowe, J.C., 2003. Surface-based transport model for mixed-size sediment. J. Hydraul. Eng. 129, 120–128. Wood, P.J., Armitrage, P.D., 1997. Biological effects of ﬁne sediment in the lotic environment. Environ. Manage. 21, 203–213. Yorke, T.H., Herb, W.J., 1978. Effects of urbanization on streamﬂow and sediment transport in the Rock Creek and Anacostia River Basins, Montgomery County, Maryland, 1962–1974. US Geol. Surv. Prof. Pap. 1003, 71.

Discussion by Bob Mussetter An important assumption in the model is that the mud must ﬁll pore spaces in the bed-material matrix before it will deposit on the bed surface. This assumption has a signiﬁcant effect on predicted mud dynamics. Our experience in the Upper Colorado River system is that mud deposits only on the surface and does not inﬁltrate in signiﬁcant quantities into the bed. As a result, the mud is much more transient than would be predicted by this model. There may be differences in physical

J. Pizzuto et al.

382

characteristics between the two river systems that would cause the mud to behave differently. However, if the assumption is not correct, this part of the result may not be realistic.

Reply by the authors There are limited data available on mud accumulations in pore spaces of streams in the study area of NW Branch. However, surveys of ﬁne-grained sediments in pores have been completed along a 24-km reach of the South River in the Shenandoah Valley of Virginia (Pizzuto et al., 2006). Along South River, an ongoing study of mercury contamination has necessitated monthly sampling of ﬁnegrained sediments in a variety of environments, including within the porous gravel streambed. Sediments are sampled at the downstream ends of pools in rifﬂe-pool sequences, where deposition is likely. Samples are obtained from 12 locations along the study reach. At each site, samples are taken from each of three 0.5 m quadrants equally spaced across the width of the stream. Pore water and sediment are pumped from the subsurface within the quadrat using a handheld Beckson bilge pump; the resulting slurry is collected into a bucket. Coarse sediment is allowed to settle, so only the silt and clay fraction is sampled. The volume of the slurry is recorded. Table 10.1 presents volumes of silt and clay sampled from the bed at 12 locations in the study area in March and April 2006. These volumetric data were converted to units of mass per unit bed area (M) using the following equation: M¼

LV ss r As

(14:22Þ

where L is the ‘‘wet’’ volume of sediment sampled (the samples listed in Table 14.3 were not dried, and therefore the volume listed includes both water and sediment), Vss the ratio of solid volume to total volume of the sample, r the density of the sediment, and As the area of the bed sampled. Values of these parameters used in the conversion are listed in Table 14.4. The average mass of silt and clay stored in the bed for all sites and both sampling dates is 0.05 kg/m2. These data suggest that considerable amounts of ﬁne-grained sediment are indeed stored in gravel pores in mid-Atlantic streams of the East Coast, U.S.A. These data, of course, are hardly comprehensive, and although subsequent monthly sampling through December, 2006 conﬁrms that these values are not unusual, much additional study is needed. Furthermore, these rather generalized observations are not sufﬁcient to conﬁrm the model used in the present simulations. Rather, they simply suggest that storage of silt and clay in pores of a gravel bed may be signiﬁcant in the region. Detailed conﬁrmation of the model used in this paper must await additional detailed study. The author is grateful for this discussion. Apparently there are important regional differences in how ﬁne-grained sediments interact with gravel streambeds that should be investigated further.

J. Pizzuto et al.

382

characteristics between the two river systems that would cause the mud to behave differently. However, if the assumption is not correct, this part of the result may not be realistic.

Reply by the authors There are limited data available on mud accumulations in pore spaces of streams in the study area of NW Branch. However, surveys of ﬁne-grained sediments in pores have been completed along a 24-km reach of the South River in the Shenandoah Valley of Virginia (Pizzuto et al., 2006). Along South River, an ongoing study of mercury contamination has necessitated monthly sampling of ﬁnegrained sediments in a variety of environments, including within the porous gravel streambed. Sediments are sampled at the downstream ends of pools in rifﬂe-pool sequences, where deposition is likely. Samples are obtained from 12 locations along the study reach. At each site, samples are taken from each of three 0.5 m quadrants equally spaced across the width of the stream. Pore water and sediment are pumped from the subsurface within the quadrat using a handheld Beckson bilge pump; the resulting slurry is collected into a bucket. Coarse sediment is allowed to settle, so only the silt and clay fraction is sampled. The volume of the slurry is recorded. Table 10.1 presents volumes of silt and clay sampled from the bed at 12 locations in the study area in March and April 2006. These volumetric data were converted to units of mass per unit bed area (M) using the following equation: M¼

LV ss r As

(14:22Þ

where L is the ‘‘wet’’ volume of sediment sampled (the samples listed in Table 14.3 were not dried, and therefore the volume listed includes both water and sediment), Vss the ratio of solid volume to total volume of the sample, r the density of the sediment, and As the area of the bed sampled. Values of these parameters used in the conversion are listed in Table 14.4. The average mass of silt and clay stored in the bed for all sites and both sampling dates is 0.05 kg/m2. These data suggest that considerable amounts of ﬁne-grained sediment are indeed stored in gravel pores in mid-Atlantic streams of the East Coast, U.S.A. These data, of course, are hardly comprehensive, and although subsequent monthly sampling through December, 2006 conﬁrms that these values are not unusual, much additional study is needed. Furthermore, these rather generalized observations are not sufﬁcient to conﬁrm the model used in the present simulations. Rather, they simply suggest that storage of silt and clay in pores of a gravel bed may be signiﬁcant in the region. Detailed conﬁrmation of the model used in this paper must await additional detailed study. The author is grateful for this discussion. Apparently there are important regional differences in how ﬁne-grained sediments interact with gravel streambeds that should be investigated further.

Two model scenarios illustrating the effects of land use and climate change

383

Table 14.3. Volume and mass per unit bed area of silt and clay stored in the bed of South River at 12 locations sampled in March and April 2006 (Pizzuto et al., 2006). Relative river mile 0.6 2 3 4.2 5.2 7.1 8.7 11.8 13.1 14.6 19 22.4

Mass/bed area (kg/m2)

Silt/clay volume (L) March, 2006

April, 2006

March, 2006

4 4 4 3 2.5 4.5 1.8 6 1 1 1.8 1.3

0.5 0.100 1.5 0.100 2.5 0.100 1.4 0.075 2 0.063 1.5 0.113 1.0 0.044 4 0.150 1.1 0.025 0.7 0.025 1.7 0.045 0.8 0.031 Mean for March and April

April, 2006 0.013 0.038 0.063 0.035 0.050 0.038 0.025 0.100 0.028 0.018 0.043 0.019 0.050

Table 14.4.

Parameters used to convert volumes in Table 14.3 to mass per unit bed area.

Parameter

Value

Volume sampled (gallons) Bed area sampled (ft2) Slurry solid volume ratio Unit weight of sediment (kg/m3)

15 36 0.5 1800

Discussion by J.P. Martin Vide The case study involves changes in the basin land use. It seems that two of them are outstanding: urbanization, which brings an increase in runoff (as pointed out by the author) and, second, a decrease in agricultural land, which probably brings a decrease in sediment supply and sediment transport. According to very basic models of river response, like the Lane’s balance analogy, from an increase in runoff and a decrease in sediment load it should result in a clear trend to river incision. In some case studies in NE Spain, where agricultural land was rapidly transformed into urban, that was a very evident consequence with large river degradation in the lapse of few decades. It was surprising to me that this was not the case in your streams.

Reply by the authors The balance between sediment transport capacity, supply, and adjustment of the stream attributed to Lane that provides the basis for this comment applies to alluvial streams that are free to adjust their morphology. However, a variety of ﬁeld

Two model scenarios illustrating the effects of land use and climate change

383

Table 14.3. Volume and mass per unit bed area of silt and clay stored in the bed of South River at 12 locations sampled in March and April 2006 (Pizzuto et al., 2006). Relative river mile 0.6 2 3 4.2 5.2 7.1 8.7 11.8 13.1 14.6 19 22.4

Mass/bed area (kg/m2)

Silt/clay volume (L) March, 2006

April, 2006

March, 2006

4 4 4 3 2.5 4.5 1.8 6 1 1 1.8 1.3

0.5 0.100 1.5 0.100 2.5 0.100 1.4 0.075 2 0.063 1.5 0.113 1.0 0.044 4 0.150 1.1 0.025 0.7 0.025 1.7 0.045 0.8 0.031 Mean for March and April

April, 2006 0.013 0.038 0.063 0.035 0.050 0.038 0.025 0.100 0.028 0.018 0.043 0.019 0.050

Table 14.4.

Parameters used to convert volumes in Table 14.3 to mass per unit bed area.

Parameter

Value

Volume sampled (gallons) Bed area sampled (ft2) Slurry solid volume ratio Unit weight of sediment (kg/m3)

15 36 0.5 1800

Discussion by J.P. Martin Vide The case study involves changes in the basin land use. It seems that two of them are outstanding: urbanization, which brings an increase in runoff (as pointed out by the author) and, second, a decrease in agricultural land, which probably brings a decrease in sediment supply and sediment transport. According to very basic models of river response, like the Lane’s balance analogy, from an increase in runoff and a decrease in sediment load it should result in a clear trend to river incision. In some case studies in NE Spain, where agricultural land was rapidly transformed into urban, that was a very evident consequence with large river degradation in the lapse of few decades. It was surprising to me that this was not the case in your streams.

Reply by the authors The balance between sediment transport capacity, supply, and adjustment of the stream attributed to Lane that provides the basis for this comment applies to alluvial streams that are free to adjust their morphology. However, a variety of ﬁeld

Two model scenarios illustrating the effects of land use and climate change

383

Table 14.3. Volume and mass per unit bed area of silt and clay stored in the bed of South River at 12 locations sampled in March and April 2006 (Pizzuto et al., 2006). Relative river mile 0.6 2 3 4.2 5.2 7.1 8.7 11.8 13.1 14.6 19 22.4

Mass/bed area (kg/m2)

Silt/clay volume (L) March, 2006

April, 2006

March, 2006

4 4 4 3 2.5 4.5 1.8 6 1 1 1.8 1.3

0.5 0.100 1.5 0.100 2.5 0.100 1.4 0.075 2 0.063 1.5 0.113 1.0 0.044 4 0.150 1.1 0.025 0.7 0.025 1.7 0.045 0.8 0.031 Mean for March and April

April, 2006 0.013 0.038 0.063 0.035 0.050 0.038 0.025 0.100 0.028 0.018 0.043 0.019 0.050

Table 14.4.

Parameters used to convert volumes in Table 14.3 to mass per unit bed area.

Parameter

Value

Volume sampled (gallons) Bed area sampled (ft2) Slurry solid volume ratio Unit weight of sediment (kg/m3)

15 36 0.5 1800

Discussion by J.P. Martin Vide The case study involves changes in the basin land use. It seems that two of them are outstanding: urbanization, which brings an increase in runoff (as pointed out by the author) and, second, a decrease in agricultural land, which probably brings a decrease in sediment supply and sediment transport. According to very basic models of river response, like the Lane’s balance analogy, from an increase in runoff and a decrease in sediment load it should result in a clear trend to river incision. In some case studies in NE Spain, where agricultural land was rapidly transformed into urban, that was a very evident consequence with large river degradation in the lapse of few decades. It was surprising to me that this was not the case in your streams.

Reply by the authors The balance between sediment transport capacity, supply, and adjustment of the stream attributed to Lane that provides the basis for this comment applies to alluvial streams that are free to adjust their morphology. However, a variety of ﬁeld

384

J. Pizzuto et al.

observations suggest that streams of the study area are strongly inﬂuenced by bedrock exposures in the streambed (Skalak, 2004), and by riparian vegetation that armors river banks (Allmendinger et al., 2005; Pizzuto and Meckelnburg, 1989). These streams are likely not entirely free to adjust their morphology, and should not be considered as traditional ‘‘alluvial’’ channels, but rather as channels that exhibit features of both bedrock and alluvial streams. For example, while urbanization has resulted in extensive incision in some areas (Trimble, 1997; Booth, 1990), many streams in mid-Atlantic urbanized watersheds have not incised deeply, and in many cases, incision cannot be demonstrated to have occurred at all (Hammer, 1972; Pizzuto et al., 2000). The model described in the paper includes this inﬂuence by representing bedrock explicitly as a dynamic component in the active layer of the bed. The author appreciates the opportunity to explain this aspect of the model in greater detail.

Discussion by Colin D. Rennie Coulthard et al. and Pizzuto et al. have tackled the important issue of the impact of climate change on river channels. River managers are increasingly concerned with climate change, and require informed estimates of likely changes in river channels (e.g., Ashmore and Church 2001). Coulthard et al. employed a sophisticated cellular landscape evolution model (CEASAR), which calculates spatiotemporally distributed sediment delivery and transport, to understand channel changes in time with and without climate change. Pizzuto et al. employed a relatively simple sediment transport model for the same ends. They argued that a simple model was justiﬁed because data and theory were not available for a more complex model. Coulthard et al. describe well the limitations of empirical approaches to understanding climate change impacts on rivers. The existing record of ﬂuvial measurements is too short to demonstrate morphological change due to climate change, and the record is convolved with channel changes due to changing land use. The paleolimnological record from stratigraphy is too coarse and uncertain to demonstrate climate change impacts. As a consequence, numerical modeling has been employed to evaluate the inﬂuence of climate change on channels. However, numerical models require validation and calibration with measured data if they are to generate reliable predictions. The data problems that limit empirical approaches also limit validation of numerical models, as acknowledged by Coulthard et al., which renders suspect the numerical model outcomes. The model results are highly uncertain, even with respect to observed trends, thus models may not even serve as exploratory tools. This is a difﬁcult dilemma, which the gravel-bed rivers community should begin to address. Gordon Grant suggested hindcasting numerical models to compare with the sedimentological record, which may be a good ﬁrst step. Reference Ashmore, P., Church, M., 2001. The impact of climate change on rivers and river processes in Canada, Geol. Surv. Can., Bull. 555, 58pp.

384

J. Pizzuto et al.

observations suggest that streams of the study area are strongly inﬂuenced by bedrock exposures in the streambed (Skalak, 2004), and by riparian vegetation that armors river banks (Allmendinger et al., 2005; Pizzuto and Meckelnburg, 1989). These streams are likely not entirely free to adjust their morphology, and should not be considered as traditional ‘‘alluvial’’ channels, but rather as channels that exhibit features of both bedrock and alluvial streams. For example, while urbanization has resulted in extensive incision in some areas (Trimble, 1997; Booth, 1990), many streams in mid-Atlantic urbanized watersheds have not incised deeply, and in many cases, incision cannot be demonstrated to have occurred at all (Hammer, 1972; Pizzuto et al., 2000). The model described in the paper includes this inﬂuence by representing bedrock explicitly as a dynamic component in the active layer of the bed. The author appreciates the opportunity to explain this aspect of the model in greater detail.

Discussion by Colin D. Rennie Coulthard et al. and Pizzuto et al. have tackled the important issue of the impact of climate change on river channels. River managers are increasingly concerned with climate change, and require informed estimates of likely changes in river channels (e.g., Ashmore and Church 2001). Coulthard et al. employed a sophisticated cellular landscape evolution model (CEASAR), which calculates spatiotemporally distributed sediment delivery and transport, to understand channel changes in time with and without climate change. Pizzuto et al. employed a relatively simple sediment transport model for the same ends. They argued that a simple model was justiﬁed because data and theory were not available for a more complex model. Coulthard et al. describe well the limitations of empirical approaches to understanding climate change impacts on rivers. The existing record of ﬂuvial measurements is too short to demonstrate morphological change due to climate change, and the record is convolved with channel changes due to changing land use. The paleolimnological record from stratigraphy is too coarse and uncertain to demonstrate climate change impacts. As a consequence, numerical modeling has been employed to evaluate the inﬂuence of climate change on channels. However, numerical models require validation and calibration with measured data if they are to generate reliable predictions. The data problems that limit empirical approaches also limit validation of numerical models, as acknowledged by Coulthard et al., which renders suspect the numerical model outcomes. The model results are highly uncertain, even with respect to observed trends, thus models may not even serve as exploratory tools. This is a difﬁcult dilemma, which the gravel-bed rivers community should begin to address. Gordon Grant suggested hindcasting numerical models to compare with the sedimentological record, which may be a good ﬁrst step. Reference Ashmore, P., Church, M., 2001. The impact of climate change on rivers and river processes in Canada, Geol. Surv. Can., Bull. 555, 58pp.

Two model scenarios illustrating the effects of land use and climate change

385

Reply by the authors This discussion raises very important issues related to our ability to forecast the inﬂuence of climate and land use changes on rivers. The discussion notes that historical, stratigraphic studies lack detail, while modeling studies are generally poorly constrained. Our study was motivated by the need to provide detailed geomorphic information that could be used to model the inﬂuence of climate and land-use changes on ﬁsh populations in headwater streams of our study area. This goal required us to predict the grain-size distribution of the bed, the ﬁne-grained sediment content of the pores, the turbidity caused by suspended sediment, and ﬂow velocities caused by a variety of discharges. While many of the processes included in the model could be veriﬁed by experimental or ﬁeld observations, the boundary conditions required by the model that relate land use and climate changes to sediment supply and discharge are virtually untestable in signiﬁcant detail, either by stratigraphic study or any other approach. The best one could really hope for is an adaptive modeling approach where model predictions are periodically adjusted after comparing model predictions to long-term monitoring of changes observed in the ﬁeld as they occur over decades. The discussion implicitly questions the value of modeling long-term geomorphic processes inﬂuenced by changes in climate and land use. The author answers such questions by explicitly noting the difference between scientiﬁc goals and the necessity of making management decisions. The goal of science is to achieve an understanding of nature that is supported by observation. This goal almost by deﬁnition cannot be achieved by modeling alone, particularly when models are applied to future conditions. In the author’s view, the present study (and others like it) represents an extension of traditional scientiﬁc goals to provide ‘‘best guess’’ working hypotheses to managers for decision-making purposes. Scientists always have the obligation, of course, of including realistic evaluation of model uncertainty when providing results such as these. In the discussion, it is suggested that models might be best evaluated by reproducing historical events. As noted, past changes to ﬂuvial systems are accessible, at least in outline form, though stratigraphic studies. This approach indeed provides some of our best current opportunities to evaluate models of long-term channel changes. Physical modeling studies can also prove useful in some instances. Hopefully, though, some selected areas will be chosen for long-term observational studies that can provide the ﬁeld data needed to evaluate future predictive models.

Discussion by Ian Reid Pizzuto et al. provide an interesting glimpse of the product of a sediment routing model. However, in adopting a modelling approach with its inevitable simpliﬁcation of processes, the general question arises as to what we ignore, or what we can afford to ignore, of the processes that have been investigated and understood in the last few decades of advancing ﬂuvial hydraulics? Illustrating this with reference to a speciﬁc

Two model scenarios illustrating the effects of land use and climate change

385

Reply by the authors This discussion raises very important issues related to our ability to forecast the inﬂuence of climate and land use changes on rivers. The discussion notes that historical, stratigraphic studies lack detail, while modeling studies are generally poorly constrained. Our study was motivated by the need to provide detailed geomorphic information that could be used to model the inﬂuence of climate and land-use changes on ﬁsh populations in headwater streams of our study area. This goal required us to predict the grain-size distribution of the bed, the ﬁne-grained sediment content of the pores, the turbidity caused by suspended sediment, and ﬂow velocities caused by a variety of discharges. While many of the processes included in the model could be veriﬁed by experimental or ﬁeld observations, the boundary conditions required by the model that relate land use and climate changes to sediment supply and discharge are virtually untestable in signiﬁcant detail, either by stratigraphic study or any other approach. The best one could really hope for is an adaptive modeling approach where model predictions are periodically adjusted after comparing model predictions to long-term monitoring of changes observed in the ﬁeld as they occur over decades. The discussion implicitly questions the value of modeling long-term geomorphic processes inﬂuenced by changes in climate and land use. The author answers such questions by explicitly noting the difference between scientiﬁc goals and the necessity of making management decisions. The goal of science is to achieve an understanding of nature that is supported by observation. This goal almost by deﬁnition cannot be achieved by modeling alone, particularly when models are applied to future conditions. In the author’s view, the present study (and others like it) represents an extension of traditional scientiﬁc goals to provide ‘‘best guess’’ working hypotheses to managers for decision-making purposes. Scientists always have the obligation, of course, of including realistic evaluation of model uncertainty when providing results such as these. In the discussion, it is suggested that models might be best evaluated by reproducing historical events. As noted, past changes to ﬂuvial systems are accessible, at least in outline form, though stratigraphic studies. This approach indeed provides some of our best current opportunities to evaluate models of long-term channel changes. Physical modeling studies can also prove useful in some instances. Hopefully, though, some selected areas will be chosen for long-term observational studies that can provide the ﬁeld data needed to evaluate future predictive models.

Discussion by Ian Reid Pizzuto et al. provide an interesting glimpse of the product of a sediment routing model. However, in adopting a modelling approach with its inevitable simpliﬁcation of processes, the general question arises as to what we ignore, or what we can afford to ignore, of the processes that have been investigated and understood in the last few decades of advancing ﬂuvial hydraulics? Illustrating this with reference to a speciﬁc

386

J. Pizzuto et al.

component of Pizzuto et al.’s model – one that is especially important to the aquatic ecologists who are driving the study – we can ask if exchanges between bed storage and the stream of the ﬁne-grained sediment which forms the interstitial matrix of most gravel-bed rivers needs a reﬁned sub-routine, given that there is probably sufﬁcient information relating the amount and vertical disposition of interstitial ﬁnes to framework gravel size parameters and ﬂow (e.g., Carling and Reader, 1982; Frostick et al., 1984; Lisle, 1989). So, should Pizzuto et al.’s ‘mud transport model’ subroutine be governed by a conditional step that ﬁrst asks for the size distributions of the bed material (surface and subsurface) deposited by the previous ﬂood before determining either inﬁltration or winnowing of ﬁnes? or do we presume a constant pre-ﬂood bed state? Furthermore, do we, as yet, have sufﬁcient knowledge to typecast the interstitial storage according to bed meso-forms (bars etc.) and, therefore, add another spatial dimension? Or is this a further challenge for that part of the community engaged in ﬁeld analyses?

References Carling, P.A., Reader, N.A., 1982. Structure, composition and bulk properties of upland stream gravels. Earth Surf. Process. Landf. 7, 349–366. Frostick, L.E., Lucas, P.M., Reid, I., 1984. The inﬁltration of ﬁne matrices into coarse-grained alluvial sediments and its implications for stratigraphical interpretation. J. Geol. Soc. London 141, 955–965. Lisle, T., 1989. Sediment transport and resulting deposition in spawning gravels, North Coastal California. Water Resour. Res. 25, 1303–1319.

Reply by the authors The author appreciates the opportunity to clarify the calculation procedure used in the model. The grain-size distribution of the bed is updated at each time step, before any attempt is made to compute the silt–clay content of the gravel streambed. Thus, the model computations do reﬂect and hopefully account for changes in the texture of the streambed through time. This does not mean, of course, that the computations are actually correct or even realistic in detail, which could only be established through a detailed observational study in the ﬁeld or laboratory. As the discussion indicates, no attempt has been made to explicitly account for mesoforms such as bars or pools and rifﬂes. The detailed morphology of the channel wetted perimeter is certainly very likely to inﬂuence the extent of storage of silt and clay in gravel-bed rivers. Indeed, ﬁeld studies by our research group on a gravel-bed river in Virginia, U.S.A., implicate large woody debris accumulations in pools as the primary control on the extent of ﬁne-grained sediment storage at this ﬁeld site (Skalak and Pizzuto, 2006; Pizzuto et al., 2006). However, knowledge of ‘‘mesoscale’’ storage processes does not appear adequate at present to allow creation of a numerical model of these processes.

386

J. Pizzuto et al.

component of Pizzuto et al.’s model – one that is especially important to the aquatic ecologists who are driving the study – we can ask if exchanges between bed storage and the stream of the ﬁne-grained sediment which forms the interstitial matrix of most gravel-bed rivers needs a reﬁned sub-routine, given that there is probably sufﬁcient information relating the amount and vertical disposition of interstitial ﬁnes to framework gravel size parameters and ﬂow (e.g., Carling and Reader, 1982; Frostick et al., 1984; Lisle, 1989). So, should Pizzuto et al.’s ‘mud transport model’ subroutine be governed by a conditional step that ﬁrst asks for the size distributions of the bed material (surface and subsurface) deposited by the previous ﬂood before determining either inﬁltration or winnowing of ﬁnes? or do we presume a constant pre-ﬂood bed state? Furthermore, do we, as yet, have sufﬁcient knowledge to typecast the interstitial storage according to bed meso-forms (bars etc.) and, therefore, add another spatial dimension? Or is this a further challenge for that part of the community engaged in ﬁeld analyses?

References Carling, P.A., Reader, N.A., 1982. Structure, composition and bulk properties of upland stream gravels. Earth Surf. Process. Landf. 7, 349–366. Frostick, L.E., Lucas, P.M., Reid, I., 1984. The inﬁltration of ﬁne matrices into coarse-grained alluvial sediments and its implications for stratigraphical interpretation. J. Geol. Soc. London 141, 955–965. Lisle, T., 1989. Sediment transport and resulting deposition in spawning gravels, North Coastal California. Water Resour. Res. 25, 1303–1319.

Reply by the authors The author appreciates the opportunity to clarify the calculation procedure used in the model. The grain-size distribution of the bed is updated at each time step, before any attempt is made to compute the silt–clay content of the gravel streambed. Thus, the model computations do reﬂect and hopefully account for changes in the texture of the streambed through time. This does not mean, of course, that the computations are actually correct or even realistic in detail, which could only be established through a detailed observational study in the ﬁeld or laboratory. As the discussion indicates, no attempt has been made to explicitly account for mesoforms such as bars or pools and rifﬂes. The detailed morphology of the channel wetted perimeter is certainly very likely to inﬂuence the extent of storage of silt and clay in gravel-bed rivers. Indeed, ﬁeld studies by our research group on a gravel-bed river in Virginia, U.S.A., implicate large woody debris accumulations in pools as the primary control on the extent of ﬁne-grained sediment storage at this ﬁeld site (Skalak and Pizzuto, 2006; Pizzuto et al., 2006). However, knowledge of ‘‘mesoscale’’ storage processes does not appear adequate at present to allow creation of a numerical model of these processes.

Gravel-Bed Rivers VI: From Process Understanding to River Restoration H. Habersack, H. Pie´gay, M. Rinaldi, Editors r 2008 Elsevier B.V. All rights reserved.

387

15 Spatial and temporal variability in stream sediment loads using examples from the Gros Ventre Range, Wyoming, USA Sandra E. Ryan and Mark K. Dixon

Abstract Sediment transport rates (dissolved, suspended, and bedload) measured over the course of several years are reported for two streams in the Gros Ventre Mountain range in western Wyoming, USA: Little Granite and Cache Creeks. Both streams drain watersheds that are in relatively pristine environments. The sites are about 20 km apart, have runoff dominated by snowmelt and are underlain by a similar geological setting, suggesting that sediment supply and rates of transport in the two watersheds may be comparable. Yet, estimated sediment yields for the two sites appear substantially different. On average, sediment load per unit watershed area was about 40% greater at Little Granite Creek than Cache Creek, with larger differences during wetter years. Moreover, while there were differences for all components of the sediment load, suspended, and bedload fractions showed the most noteworthy contrast where nearly three times more material was exported from Little Granite Creek on an annual basis. Speculatively, this is attributed to contributions of sediment from several chronic sources in the Little Granite Creek watershed. Similar to other studies of sediment transport in gravel-bed streams, the range of measured bedload and suspended sediment in this study were quite variable. An assessment of annual differences in the 13 years of bedload record for Little Granite Creek indicated that variability could not be ascribed to between-year differences. Instead, the data appear to represent the same population of highly variable transport rates. However, the inability to distinguish between years could be due to the infrequency with which data were collected each year. Seasonal variability was evident in the suspended sediment record of Little Granite Creek where there were higher rates on the rising limb of the snowmelt hydrograph, indicative of a ﬂushing of sediment with the onset of snowmelt. Baseline data on rates of sediment transport provide useful information on the inherent variability of stream processes and may be used to assess departure due to natural or anthropogenic disturbances. In August 2000, wildﬁre burned portions of E-mail addresses: [email protected] (S.E. Ryan) ISSN: 0928-2025

DOI: 10.1016/S0928-2025(07)11134-2

S. E. Ryan, M. K. Dixon

388

the Little Granite Creek watershed, presenting an opportunity to quantify increases in sediment loads associated with large-scale disturbance. The results of 3 years of postﬁre monitoring showed substantial increases in suspended sediment transport on the rising limb of the snowmelt hydrograph and during several summer thunderstorms. Suspended sediment yields calculated for the post-ﬁre years were higher for the ﬁrst year and have decreased over time, indicating a return to baseline levels within a few years following the wildﬁre. In contrast, there were no detectable increases in the rate of bedload transport over the pre-burn values suggesting differences in the rapidity with which the two sediment components respond to disturbance.

1.

Introduction

A primary difﬁculty in generalizing sediment transport processes in mountain streams involves accounting for spatial and temporal variability in sediment supply common to these systems. On a large scale, both sediment supply and stream sediment loads in upland watersheds are inﬂuenced by a number of factors, including topography, geology, vegetation, land management, and natural disturbances (Walling, 1988). Given this, similarities in rates and processes might be inferred across regions inﬂuenced by similar geologic and climatic factors. Yet, while similarities in processes for streams draining the same geographic region often occur, interactions occurring at smaller scales may well create conditions unique to an individual watershed (Walling and Webb, 1982). In complex terrain, there are often large differences in lithology, hillslope processes, and erosion rates occurring over short distances, even within the same watershed, that can affect sediment transport processes in channels downstream. Total sediment export from a watershed occurs as dissolved load, in suspension, and as bedload. Much of the attention on sediment loads in channels draining steepland environments is given to coarse sediment that is transported sporadically along the channel bottom as bedload (Ryan and Dixon, 2002). This emphasis is given because coarse sediment comprises the lag deposit in which the channel conduit is formed and is the template for many aquatic functions. The cutoff between sediment moved as bedload and suspended load is somewhat arbitrary and overlapping. Depending on current velocity and turbulence, sediment sizes moved as bedload under moderate ﬂow may be moved in suspension during increased discharge. Though arguably the processes are related, the two components are treated separately based on the differing methods used to collect suspended or bedload materials. Moreover, dissolved and suspended loads often comprise a greater proportion of the total sediment load of steepland streams (Dietrich and Dunne, 1978; Alvera and Garcia-Ruiz, 2000). While less important in forming the physical structure of coarsegrained channels (Parker, 1978a,b), suspended and dissolved loads have biological signiﬁcance (Wohl, 2000). For instance, suspended sediment concentration (SSC) affects light available for photosynthesis in aquatic plants and excessive inﬁltration of ﬁnes into the channel substrate poses hazards to ﬁsh and macroinvertebrates in hyporheic environments (Waters, 1995). Hence, all three components should be addressed in assessing sedimentation effects on channel form and aquatic function.

Spatial and temporal variability in stream sediment loads

389

In order to gain improved understanding of stream processes and baseline sediment yields, a series of ﬁeld studies were carried out in two watersheds near Jackson, Wyoming, USA. The original intent of the studies varied, but the result was a compilation of an extensive database on dissolved, suspended, and bedload transport rates that can be used to develop estimates of sediment yield with a relatively high level of conﬁdence. Little Granite Creek and Cache Creek are both located in the Gros Ventre Mountain range near Jackson Hole and are upland contributors to the Snake River. The geology of the Gros Ventre range consists primarily of Paleozoic and Mesozoic sedimentary rock formations, many of marine origin. The formations are deformed and largely unstable; parts of the Gros Ventre range are characterized by some of the highest densities of landslides in the United States (Wyoming State Geological Survey, 2001). The area is also inﬂuenced by active faulting and a relatively high risk of earthquakes. Hence, sediment delivery to stream systems from an unstable landscape is comparatively high for many of the watersheds in this area. However, despite landscape scale similarities common to the Gros Ventre range, there are substantial differences in measured sediment loads from the Little Granite and Cache Creek watersheds, suggesting variation in sediment delivery and transport within a province of complex terrain. The overall goal of this paper is to compare and contrast sediment loads from two proximal watersheds in an effort to elucidate processes occurring at local scales and the factors that control them. Speciﬁcally, three topic areas are addressed in this paper. First, the rating curves developed for each of the three sediment components are presented and the records are assessed for annual and seasonal variations. Second, estimates of annual sediment yield are compared and differences are tentatively linked to overall patterns of mass wasting in the two watersheds. Third, the effects of disturbance on the rates and patterns of instream sedimentation are evaluated for the Little Granite Creek watershed, portions of which were burned by wildﬁre in August 2000.

2.

Study sites

Little Granite Creek and Cache Creek are two small mountain watersheds located on the Bridger-Teton National Forest (Fig. 15.1). Cache Creek drains a 27.5 km2 area of steep terrain and inner gorges largely under mixed subalpine forest and meadow vegetation cover. Elevations range from 2057 m at the gauge site to 3141 m at Cache Peak. The basin is a long, oblique shaped watershed, oriented in a northwest direction. The Cache Creek site is at a United States Geological Survey (USGS) gauge (13018300) and is part of the Hydrologic Benchmark Network (HBN) established by the USGS in 1958. The original goal of the HBN program was to establish a network of gauges located at pristine sites to provide a baseline against which the inﬂuence of environmental factors on streamﬂow and water chemistry could be evaluated (Cobb and Biesecker, 1971). The program was active for over 30 years at some locations. The USGS established the recording gauge station at Cache Creek in 1963 and the gauge has been in continuous operation since then. Water quality monitoring, initiated in 1965, was discontinued at Cache Creek in 1995 due to declining funding. During the

390

S. E. Ryan, M. K. Dixon

Figure 15.1. Location of Cache and Little Granite Creek watersheds in western Wyoming, USA. Dark dashed line indicates boundary of the Gros Ventre Wilderness area located to the north of the line. Stipple pattern indicates area administered by the Bridger-Teton National Forest.

active period, some water quality measurements were taken on a monthly basis while others were taken bi-monthly, semi-annually, or annually. Little Granite Creek drains a 54.6 km2 area near the site of a former USGS gauging station (13019438) that was active between 1982 and 1993. The sediment sampling program at this site began in 1982 as part of environmental monitoring in conjunction with planned exploration and extraction of fossil fuels in the upper basin. Though the exploratory effort was eventually abandoned, sediment monitoring was continued by the USGS through 1993. The database was expanded when additional bedload samples were collected by US Forest Service (USFS) personnel during high ﬂow in 1997 (Ryan and Emmett, 2002). Basin elevation ranges from 3329 m at an unnamed high point to 1948 m at the conﬂuence of Little Granite and Granite Creeks. The basin is rounded in shape and much of the valley area upstream of the sampling site is characterized by relatively open valley bottoms with mixed subalpine forest and meadow vegetative cover. The watershed is oriented toward the south to southeast. Upper Little Granite and Boulder Creeks, primary tributaries to Little Granite Creek also discussed in this paper, drain 19.7 and 20.7 km2 areas, respectively (Fig. 15.1). Boulder Creek was burned in August 2000 and Upper Little Granite Creek served as a control watershed in a study on the effects wildﬁre on stream sedimentation (Ryan et al., 2003).

Spatial and temporal variability in stream sediment loads 2.1.

391

Geology

The geology of the Cache Creek basin is varied but consists primarily (60%) of deformed sedimentary rock formations ranging in age from Paleozoic to Cretaceous including: limestone, dolomite, shale, siltstone, sandstone (Love and Christiansen, 1985). Pleistocene or Pliocene conglomerates (15%) are also present and include clasts of limestone in a partially lithiﬁed carbonate matrix. The main channel cuts through this unit in a narrow inner gorge above the gauging station. Unconsolidated units, consisting of alluvium, colluvium, Quaternary landslide debris, and a large slump block, make up about 23% of the surﬁcial area. A small portion of upper Cache Creek was glaciated and these surfaces are largely stabilized by alpine meadow and lodgepole/subalpine forest cover. The geology underlying Little Granite Creek is also varied and composed primarily of deformed sedimentary rocks of Cretaceous and Tertiary age, including conglomerate, sandstone, claystone, and limestone (Love and Christiansen, 1985). The upper portion of the basin was glaciated and is mantled by till and glacial outwash, likely correlative of Bull Lake and Pinedale aged deposits (as generalized in Leopold and Emmett, 1997). At high elevations, these surfaces are largely devoid of vegetation cover and actively contribute sediment to the channels, particularly during snowmelt runoff. Primary sources of sediment at both sites are from mass wasting, including active earthﬂows from unstable hillslopes and slumping from undercut terraces and road cuts. Scour from the channel bed and banks also contribute substantially to the sediment load. However, the type of landslides and recency of movement are quite different in the two watersheds, with substantially more active instabilities occurring in Little Granite Creek. Areas affected by landslides were identiﬁed using landslide maps generated by the Wyoming State Geological Survey (2004) overlain on topographic coverage. Relative activity and type of slide was further delineated using aerial photography of the watersheds taken in 1994 and 2001 and through ﬁeld reconnaissance in 2004.

2.2.

Climate

While no permanent climatic station exists within either watershed, there are three stations in the vicinity from which climatic conditions may be inferred. These include a station in the town of Jackson (elevation 1900 m), the Granite Creek automated SNOTEL (snow telemetry) station (elevation of 2063 m), and the Phillips Bench SNOTEL (elevation 2500 m) (Western Regional Climate Center, 2003). Snow cover is typically present at all stations between November and March. Precipitation at the lower station averages 400 mm per year and is fairly well-distributed between seasons. Mean annual precipitation at the middle elevation is about 800 mm annually, with the majority of precipitation occurring between November and January in the form of snow. Mean annual precipitation at the higher elevation is about 1120 mm annually, with snow accumulating between November and May. The mean annual temperature at the Jackson site is 3.31C, 1.51C at the Granite Creek site and 1.31C at the Phillips Bench site.

S. E. Ryan, M. K. Dixon

392 2.3.

Runoff patterns

Runoff at both sites is generated primarily by the melting of snow in the spring, with peak ﬂows occurring between mid-May and mid-June. Flow typically peaks earlier in the year at Little Granite Creek (late May) than at Cache Creek (early June) due to the differences in aspects between the watersheds. While thunderstorms are common in the summer, they typically produce only small rises in the hydrograph. The 1.5-year return interval ﬂow (a surrogate for the bankfull discharge) at Cache Creek is 1.7 m3 s1, as estimated using Log Pearson III analysis on the annual maximum ﬂood series for the period 1962–1998 (US Interagency Advisory Committee on Water Data, 1982). The 95% conﬁdence limits on this estimate are 1.44 and 1.97 m3 s1. The calculated 1.5-year return interval ﬂow at Little Granite is 5.95 m3 s1 and the 95% conﬁdence limits on this estimate are 3.71 and 8.38 m3 s1. The conﬁdence bands on the estimates are relatively wide due to uncertainty associated with a limited (10 years) record of ﬂow (Linsley et al., 1975). 2.4.

Land use

The primary former land use in Cache Creek consisted of two small coal mines that provided coal for local needs; these were abandoned in the 1930s. Small-scale logging and grazing by cattle also occurred prior to 1940 (Clark et al., 2000) but the basin was closed to these activities to protect the primary water supply to the town of Jackson. The upper basin is in a wilderness area (Fig. 15.1) and is nearly pristine. Outside of the wilderness boundary, an old roadbed follows the course of Cache Creek approximately 4.5 km upstream from the gauging station. Though closed to motorized trafﬁc beginning in the 1960s, the roadway and trail network is very heavily used for summer and winter recreation. At a few locations, the eroding roadbed impinges on the stream channel and contributes sediment to Cache Creek, though these sources tend to be small. Upland land uses in Little Granite Creek include cattle grazing for a few months in summer and dispersed recreation (camping, hiking, horseback riding, and hunting). A gravel access road parallels the stream for about 2.4 km in the lower end of the watershed and this is undercut by the stream channel at several places. Former land uses include coal extraction and mining camps in the valley near upper Little Granite Creek. These former and present uses are small in scale and their inﬂuence on sediment delivered to the channel are relatively nominal, though the inﬂuence of the mining camps may have been more substantial when they were active. 2.5.

Sediment monitoring sites

The channel near the Cache Creek site is single thread, stable, and incised into terraces. The area of the ﬂoodplain is small and discontinuous. The active channel width ranges from 5 to 7 m and mean depth at bankfull stage is 0.40 m. Water surface slope at bankfull in the vicinity of the gauge is 0.024. The bed surface consists primarily of cobbles and coarse gravel with a few concentrations of boulders; D16 is 10 mm, D50 is

Spatial and temporal variability in stream sediment loads

393

45 mm, and D84 is 115 mm. In the subsurface, D16 is 4 mm, D50 is 20 mm, and D84 is 58 mm. The channel form alternates between areas of irregularly spaced step-pools to plane-bed morphology (as deﬁned by Montgomery and Bufﬁngton, 1997). Riparian vegetation in the vicinity of the stream consists of willows and a sedge/grass mixture with some coniferous trees. The valley bottom is densely forested by Douglas Fir. The channel in the vicinity of the Little Granite Creek study site is also single thread and plane bed, with slight sinuosity. Channel width and mean depth at bankfull are 9.8 and 0.43 m, respectively. Mean water surface slope at bankfull ﬂow is 0.020. Bed material at the surface ranges from gravel to small boulders; D16 is 24 mm, D50 is 90 mm, and D84 is 208 mm. In the subsurface, D16 is 2.5 mm, D50 is 17 mm, and D84 is 41 mm. Channel banks are largely stable and densely vegetated by willow and lodgepole pine.

3. 3.1.

Methods Bedload measurements

Samples of bedload were collected using a pressure-difference bedload sampler (Helley and Smith, 1971). The body of the sampler is constructed of 1/4-inch thick cast aluminum with an expansion ratio of 3.22 and ﬁtted with a catch bag of 0.25 mm nylon mesh. The oriﬁce through which bedload passes is 76 mm 76 mm so the sampler is capable of capturing grain sizes up to small cobble. The same type of sampler was used to collect bedload at both sites. The Helley–Smith sampler has a high trapping efﬁciency for the predominant range of bedload moved in the two watersheds (Emmett, 1980) and has been used successfully to predict total bedload yield from other small mountain streams (Ryan and Porth, 1999). Bedload was sampled using the single equal width increment (SEWI) method (Edwards and Glysson, 1998), which involves placing the sampler at equally spaced positions (about 0.5 m apart) along a cross-sectional transect. Typically, two complete traverses of the channel are made and all materials collected from both traverses are combined into a single sample that represents a spatially and temporally averaged rate of transport. The samples were oven dried and sieved using standard methods for grain size analysis (Folk, 1968). Rates of transport were calculated using the total mass of the sample divided by the total sampling time and by the width of the sampler to obtain the unit bedload transport rate (kg m1 s1). This value is multiplied by the width of the channel to obtain the mean transport rate (in kg s1) through the cross-section. Bedload measurements were taken primarily during periods of snowmelt runoff. This is because summer and fall rainstorms tend to be small and have minimal inﬂuences on water stage (and, hence, bedload transport) (Ryan and Emmett, 2002) and so they are not regularly sampled. Bedload measurements were obtained during snowmelt runoff in 1999 at Cache Creek and between 1982 and 1997 at Little Granite Creek. A total of 60 bedload samples were collected at Cache Creek at ﬂows up to 1.3 times the bankfull discharge. Although bedload data were collected for only one snowmelt period at this site, we presume that these data are representative of the period being compared. This is based on an observation of overall system stability and

S. E. Ryan, M. K. Dixon

394

lack of annual differences in the other sediment records that would suggest substantial changes in sediment supply occurring in Cache Creek. Furthermore, data were collected from a relatively wide range of ﬂows, thereby alleviating some of the difﬁculties encountered when attempting to ﬁt a model to a limited bedload dataset (as discussed in the analysis of annual variability described in Section 4.1.). At Little Granite Creek, more than 120 measurements of bedload transport were acquired at ﬂows ranging from about 0.05 bankfull (approximately 0.3 m3 s1) to nearly twice the bankfull discharge (11.3 m3 s1). Daily bedload yield for both sites was estimated from bedload rating curves (Ryan and Emmett, 2002; Ryan et al., 2005b) and USGS records of mean daily discharge. These values were summed by water year to estimate annual bedload yield for the years 1983–1992, the common period of record for these two sites. Annual yield is expressed as load (metric tonnes t) per unit contributing basin area (km2). 3.2.

Suspended sediment measurements

Suspended sediment observations at Little Granite Creek were obtained from USGS records taken between 1982 and 1993 (US Geological Survey, 2004). There were about 170 observations made during this period for ﬂows up to two times the bankfull discharge (Ryan and Emmett, 2002). USGS data collected between 1968 and 1996 at Cache Creek (US Geological Survey, 2004) were supplemented by measurements taken by USFS personnel in 1999. About 290 measurements of suspended sediment were made for this site. Samples were collected at several, equally spaced positions on the cross-section using depth integrating samplers, such as a DH-48 (Federal Interagency Sedimentation Project, 2005) while wading or from sampling bridges. Samples were ﬁltered, dried, and weighed to determine the total organic/inorganic mass of suspended sediment. SSC is expressed as the total mass of material per volume of liquid sample. Concentration was converted to transport rate (kg s1) by multiplying by instantaneous ﬂow observed at the time the sample was collected. Suspended sediment rating curves were developed from ﬂow observations and transport rates. Estimates of daily suspended sediment yield obtained using these curves were summed to determine annual yield for the suspended component. 3.3.

Data on total dissolved load

Total dissolved load was obtained from USGS records for both sites (US Geological Survey, 2004) and, for Little Granite Creek, from unpublished data from the Bridger-Teton National Forest. Altogether, 156 observations of total dissolved concentrations were made for Little Granite Creek. Samples were collected on a monthly basis between 1981 and 1984, and more frequently between 1985 and 1992. Water quality samples were collected on a near-monthly basis at Cache Creek between 1966 and 1982 (US Geological Survey, 2004). The sampling frequency was reduced to bi-monthly between 1982 and 1995. A summary of the complete set of 272 water quality samples from Cache Creek collected under the HBN program is provided in Clark et al. (2000).

Spatial and temporal variability in stream sediment loads

395

Total dissolved concentration was determined either by drying water samples at 1801C or from speciﬁc conductance measurements converted to an estimate of total dissolved concentration using a conversion factor (Stednick, 1991). Conversion factors were selected based on the single value that gave the best correspondence between measurements obtained with both methods. These were applied to observations where only speciﬁc conductance was measured. A conversion factor of 0.55 was used in estimating dissolved concentrations at Cache Creek. The conversion factor for Little Granite Creek was 0.6. Total dissolved concentration was converted to transport rate (kg s1) by multiplying by the instantaneous ﬂow observed at the time the sample was collected. Mean daily and total annual dissolved yields were estimated from rating curves, as described for the bedload and suspended load components. 3.4.

Discharge measurements

Discharge at the time samples were collected was determined either from USGS records (when available) or from ﬂow rating curves developed using direct measurements and stage readings from staff plates installed near the sampling sites. Discharge was calculated from the product of mean velocity, interval width, and total depth determined at numerous subsections, then summed for the channel cross-section (Buchanan and Somers, 1969; Nolan and Shields, 2000). Mean velocity was measured using a Price AA and/or mini-current meters at 0.6 depth.

4. 4.1.

Results Annual and seasonal variation in sediment loads

Rating curves ﬁt to measurements of sediment transport typically exhibit considerable variability about the curve (Fig. 15.2). The range of measured bedload values for a given discharge can vary by one or more orders of magnitude, indicating that stream ﬂow alone is not the only factor affecting rates of transport. Both short- and longterm ﬂuctuations in sediment supply, including those occurring at seasonal and annual timescales, can contribute to variability demonstrated in records of sediment transport (Gomez et al., 1989). While the bedload data collected at Little Granite and Cache Creeks were of insufﬁcient frequency for detecting changes in seasonal patterns (i.e., hysteresis), the 13 years of record from Little Granite Creek permitted a comparison of data collected annually. Typically, there were too few data (4–12 observations) or the data were collected over too narrow a range of ﬂow to permit the comparison of bedload models ﬁt to annual data. In previous work, an assessment of the adequacy of sample size indicated that a minimum of 20 samples were required to deﬁne a relationship (Troendle et al., 1996) and between 60 and 80 samples were required to discriminate between means in groups of bedload samples on the order of 30%, for varying levels of power (Ryan and Porth, 1999). Since there were too few samples collected each year for this level of assessment, the observations from any one year were compared with data from all other years to determine if they fell outside of

S. E. Ryan, M. K. Dixon

396 10

10 (a)

Bedload Suspended Dissolved

1 Transport Rate (kg s-1)

Transport Rate (kg s-1)

1

0.1

0.01

0.001

0.1

0.01

0.001

0.1 1 Discharge (m3 s-1)

Bankfull

Bankfull

0.0001 0.01

(b)

Bedload Dissolved Suspended Early Season Suspended

10

0.0001 0.01

0.1 1 10 Discharge (m3 s-1)

100

Figure 15.2. (a) Rating curves for Cache Creek. The form of the equation that best ﬁt the data was used, with no a priori decision as to the form of the function. The curve for bedload data is an exponential ﬁt [y ¼ a1 ðeb1 x Þ] while the curves for suspended sediment and dissolved load are power ﬁts in the form [y ¼ b2 ðxc2 Þ], where x is the discharge and y is the measured transport rate. The data are displayed on log–log scale so that the three components can be shown simultaneously. (b) Rating curves determined for Little Granite Creek. The curves are power ﬁts in the form [y ¼ b2 ðxc2 Þ]. A substantial number of large suspended samples were collected on the rising of the seasonal hydrographs, justifying two curves to better deﬁne the suspended sediment rating.

the 95% conﬁdence limits of the predicted values (Fig. 15.3). The dependent variable ‘‘transport rate’’ was log-transformed to linearize the data and improve normality and heterogeneous variance problems. In most comparisons, measured transport rates in a given year consistently fell within the 95% conﬁdence intervals determined for data collected in all other years. The years that differed to some degree were: (1) 1987 where the four measured points were close to the limits of the upper conﬁdence band (Fig. 15.3c) and (2) 1993 where 5 of the 12 points were close to the lower conﬁdence band (Fig. 15.3e). Still, values for these 2 years are comparable to those measured for other years and it is therefore likely that the data represent the same population of transport rates. In short, there is a high degree of variability in transport rates and overall variability in the dataset cannot be ascribed to annual differences in one or more years. Correspondence between annual measurements was also suggested by Leopold (1994) using a smaller portion of the same dataset. Similar to observations from the bedload dataset, there were no substantial differences in suspended sediment transport measured between years for either site (separation of data by year not shown). However, a larger number of samples permitted an assessment of seasonal effects in the suspended sediment records. Data from the rising limb of the snowmelt hydrograph were separated from the recessional limb and models were ﬁt to both groupings. Measurements of suspended sediment on the rising limb from Little Granite Creek were often twice those obtained from comparable discharges on the falling limb (Fig. 15.2b). This is likely due to early season ﬂushing of sediment

1 0 -1 -2 -3 -4 -5 0

2

4 6 8 Discharge (m3 s-1)

10

12

2 c) 1987 1 0 -1 -2 -3 -4 -5 0

2

2

4 6 8 Discharge (m3 s-1)

10

12

e) 1993

1 0 -1 -2 -3 -4 -5 0

2

4 6 8 Discharge (m3 s-1)

10

12

Log Transformed Bedload Transport

a) 1982

Log Transformed Bedload Transport

2

Log Transformed Bedload Transport

Log Transformed Bedload Transport

Log Transformed Bedload Transport

Log Transformed Bedload Transport

Spatial and temporal variability in stream sediment loads

397

2 b) 1984 1 0 -1 -2 -3 -4 -5 0

2

4 6 8 Discharge (m3 s-1)

10

12

4 6 8 Discharge (m3 s-1)

10

12

10

12

2 d) 1990 1 0 -1 -2 -3 -4 -5 0

2

2

f) 1997

1 0 -1 -2 -3 -4 -5

0

2

4

6

8

Discharge (m3 s-1)

Figure 15.3. (a–f). Bedload transport rates (log transformed) measured at Little Granite Creek for some individual years (1982, 1984, 1987, 1990, 1993, and 1997). Lines shown are the 95% conﬁdence bands on the mean (inner band) and prediction values (outer band) for all other years (open circles), exclusive of the data for the year of comparison (black circles).

associated with the onset of snowmelt runoff (e.g., Reid et al., 1985). Therefore, two rating curves were used to characterize suspended sediment transport at this site. This separation becomes important later in the assessment of the impacts of wildﬁre because there are distinctions in suspended sediment transport occurring on different parts of the hydrograph in the post-ﬁre record. Only one function was ﬁt to suspended

398

S. E. Ryan, M. K. Dixon

sediment data collected at Cache Creek because no early season difference was apparent in the data. Dissolved concentration typically decreases with increasing discharge, reﬂecting dilution in the channel. Concentrations at low ﬂows in Little Granite Creek ranged from 250 to 300 mg L1 and were about 160 mg L1 at higher discharges. In Cache Creek, there was no signiﬁcant relationship between dissolved concentration and discharge. The mean estimate over the entire range of ﬂow was 178 mg L1 (717 mg L1, standard deviation). Calculated dissolved loads are determined largely by the amount of discharge at the time of sample collection and so no separation in curves due to seasonal or annual affects is typically manifest (Fig. 15.2a and b).

4.2.

Estimates of annual sediment yield from two watersheds

Substantial differences in estimates of annual sediment yield (based on the rating curves) were calculated for Little Granite and Cache Creek sites for the common period of record (1983–1992). Typically, the estimate of total load per unit watershed area from Cache Creek was 40% lower than that estimated for Little Granite Creek. The 10-year average for Cache Creek was determined to be about 80 vs. 135 t km2 for Little Granite Creek (sum of three components in Fig. 15.4). Predicted differences between total load were more substantial in wetter years (1983, 1984, 1987) and more nominal in drier years (1987, 1990, 1992) (Fig. 15.4). Similarity in sediment loads in drier years is due to the equivalence and the high contribution of the dissolved load to the total yield – about 60 t km2 in both watersheds. The mean annual dissolved load for the 10-year period was 70 t km2 (about 80% of the total) at Cache Creek and 90 t km2 at Little Granite Creek (about 70% of the total). In contrast to dissolved loads, estimates of the suspended load and bedload components showed greater dissimilarity between the two watersheds. On an average, annual suspended load was 37 t km2 at Little Granite Creek and 13 t km2 at Cache Creek (Fig. 15.4). An average of 1.3 t km2 of bedload was estimated for Cache Creek (less than 2% of the total load) while the annual bedload yield for Little Granite Creek averaged 6 t km2 (5% of the total load). This indicates that about three times the amount of non-dissolved solids are transported per unit watershed area in Little Granite Creek relative to Cache Creek, with suspended sediment contributing the greatest portion of the non-dissolved load in both watersheds. Speculatively, dissimilarities in suspended load and bedload may be attributed to gross differences in hillslope erosion processes that deliver more coarse sediment to Little Granite Creek. An evaluation of landslides compiled by the Wyoming State Geological Survey (2001) reveals about ﬁve times the number of landslides in the Little Granite Creek watershed (118) compared to Cache Creek (23). Inspections of aerial photographs and ﬁeld reconnaissance showed that many of the slides in both watersheds are large, slow-moving, earthﬂow types that involve large portions of hillslope (Varnes, 1978). Several of the slides identiﬁed at Little Granite Creek directly impinge on stream channels. Where these slides impinge, the toeslopes are actively raveling and, in at least three cases, the channel has been pushed into the far side of the valley bottom, creating a second set of slides where the hillslope is undercut. These

Spatial and temporal variability in stream sediment loads

399

a) Cache Creek

160

Sediment Yield (t km-2)

140

Bedload Suspended Load

120

Dissolved Load

100 80 60 40 20 0 1983

1984

1985

1986

1987

1988

1989

1990

1991

b) Little Granite Creek 160

1992

10 year Mean

Post Fire Period

Sediment Yield (t km-2)

140 120 100 80 60 40 20 0 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992

10 year Mean

2001 2002 2003

Figure 15.4. Estimates of mean annual yields determined for bedload, suspended load, and dissolved load at (a) Cache and (b) Little Granite Creeks between 1983 and 1992. Data from 2001–2003 in (b) are yields estimated for the period following wildﬁre in the Little Granite Creek watershed, discussed in Section 4.3.

slides are a chronic source of ﬁne-to-coarse sediment in this basin as materials deposited in and adjacent to the channel are reactivated during high runoff. Additionally, the upper portion of the Little Granite watershed was glaciated and there are active sources of sediment from glacial and landslide debris that contribute to the relatively high suspended sediment load during snowmelt runoff. By contrast, large, chronic sources of sediment largely lack connectivity with the stream network in Cache Creek, which ﬂows through a stable inner gorge above the gauging station. The upper portion of Cache Creek was glaciated to a lesser extent than Little Granite Creek and most of these surfaces are stabilized by subalpine meadow and forest vegetation. Hence, differences in the suspended and bedload portions of these two watersheds may be linked to variation in geologic controls, patterns of mass wasting,

S. E. Ryan, M. K. Dixon

400

and the chronic contributions of sediment derived from impinging landslides and glacial debris. The link between sediment sources and loads in Little Granite Creek and Cache Creeks is an area of on-going investigation.

4.3.

Variation in sediment loads following wildfire

Natural and anthropogenic disturbances in watersheds often cause increases in sediment loads in channels downstream from the disturbed area. The net effect is typically difﬁcult to predict and depends both on the nature of the disturbance and the sequence and intensity of precipitation following the event. Often there is little baseline data against which the effects of the disturbances may be compared. In August 2000, a substantial portion of the Little Granite Creek watershed was burned by wildﬁre, presenting an opportunity to quantify increases in sediment loads due to the destruction of vegetation and loss of soil cohesion during high severity burns. Speciﬁcally, about 75% of the forest vegetation in the Boulder Creek watershed (Fig. 15.1) was moderately to severely burned. This watershed constitutes nearly 40% of the area of the Little Granite Creek watershed and so there was high potential for increased sediment loading in the years following the ﬁre. The Little Granite Creek study site was re-instrumented in 2001–2003 so that changes in timing and increases in sediment loads could be assessed. Methods and instruments similar to those used in previous years were deployed in 2001 and 2002, though only bedload and suspended sediment loads were assessed. The primary modiﬁcation in methods was the use of automated water samplers in addition to DH-48 suspended sediment samplers (complete description of methods provided in Ryan et al., 2003). The automated samplers had been programmed to collect water once every 4 h between May and October in the ﬁrst 2 years post-ﬁre, generating about 800 samples per year. In 2003, turbidity sensors were deployed and these measurements were correlated with suspended sediment samples to derive estimates of sediment concentration (Ryan et al., 2006). Because the change in methods in 2003 gave results that have greater uncertainty, only results from 2001 and 2002 are presented herein. No bedload was measured in 2003. Similar to areas throughout the western United States, snow pack and resulting stream ﬂows were relatively low in Little Granite Creek in water years 2001 and 2002 (Ryan et al., 2003). During this 2 year period, discharge exceeded bankfull for only a few hours during a prolonged rainstorm coincident with snowmelt runoff in May 2001 and during a period of rapid warming in May 2002. Later in the summers of both years, small increases in runoff occurred in response to low-to-moderate intensity thunderstorms but none exceeded bankfull. Rates of bedload transport measured during snowmelt runoff in these years showed no increase, falling within the range of pre-burn values (Fig. 15.5). This suggests that either few, new sources of coarse sediment were introduced following the burn or that there is a lag period between the introduction of new sediment and the time it takes to reach the mouth of the watershed. Based on an absence of changes in surveyed cross-sections in the burned watershed (Ryan et al., 2003) and the lack of debris-ﬂow generating landslides identiﬁed during ﬁeld reconnaissance, the former is surmised.

Spatial and temporal variability in stream sediment loads

401

10 Pre-burn 2001 2002

Bedload Transport (kg s-1)

1

0.1

0.01

0.001

0.0001 0.1

1 10 Discharge (m3 s-1)

100

Figure 15.5. Comparison of rates of bedload transport measured in 2001 and 2002 against pre-burn values (in gray). Range of values from both post-ﬁre years overlap those measured between 1982 and 1997.

In contrast, there were some substantial increases in suspended sediment loads relative to both the pre-burn period and to measurements obtained from the control site, Upper Little Granite Creek (Fig. 15.1). The magnitude of these increases varied for different parts of the runoff season (Fig. 15.6). For instance, measured rates of suspended sediment transport were about 3–5 times higher on the rising limb of the snowmelt hydrograph relative to pre-burn values, while there were no substantial differences on the falling limb. In mid summer, though low-to-moderate intensity thunderstorms (10–20 mm h1) typically raised the water stage by only a small amount, there were increases in suspended sediment between one and two orders of magnitude associated with these storms. However, they did not exceed some baseline values measured for that part of the hydrograph prior to burning.

S. E. Ryan, M. K. Dixon

402 100

100

b) 2002

a) 2001 High Concentrations Rising Limb 10

10 Rising Limb and Peaks

High Concentrations

Suspended Sediment (kg s-1)

1

Suspended Sediment (kg s-1)

Individual Storms

0.1 Falling Limb 0.01

1

0.1 Mid Summer Storms

0.01

0.001 Falling Limb and Baseflow

Baseflow

0.001

0.0001 Pre-burn 2001

Pre-burn 2002

10-5 0.01

0.1

1

Discharge (m3 s-1)

10

100

0.0001 0.01

0.1

1

10

100

Discharge (m3 s-1)

Figure 15.6. Comparison of rates of suspended sediment transport measured in (a) 2001 and (b) 2002 against pre-burn values (x). Black solid lines approximate the ranges of values measured for different parts of the hydrograph.

In late summer 2001, two larger rainstorms produced the highest SSCs measured at Little Granite Creek to date (Fig. 15.6a, High Concentrations). A brief (o 15 min) but intense rainfall (50 mm h1) on August 9 generated several ﬁne-grained, organic rich mudﬂows from gullies within the burned area. The tracks of these ﬂows could be traced across alluvial fans areas and into the stream network. SSC associated with this event was 25,000 mg L1 and it took over a week for measured concentrations to return to baseline values. By contrast, concentrations measured 280 mg L1 in the control watershed, returning to baseline within 8 h. A second series of storms on September 13 produced another substantial spike in suspended sediment. The measured peak concentration was about 48,000 mg L1 or about four orders of magnitude greater than baseline, whereas peak concentration in the control watershed was 1300 mg L1, or about the same level observed during snowmelt runoff. Concentrations measured at Lower Little Granite during these two events were in the range of hyperconcentrated ﬂow (Costa, 1984; Rickenman, 1991). The relationship between stream sediment concentrations and summer thunderstorms in 2002 differed from 2001 in that smaller increases in suspended sediment loads were associated with two moderate storms (Fig. 15.6b). The difference in response is attributed to the combined inﬂuences of: (1) lower intensity storms in 2002 and (2) vegetation regrowth on hillslopes and riparian areas (Dwire et al., 2006) reducing the supply of sediment. It may also be attributed to increased sediment

Spatial and temporal variability in stream sediment loads

403

storage in the channel due to: (1) tree fall in the reach above the study site and (2) new beaver ponds constructed in the interim period. In the period between 2001 and 2003, calculated annual suspended sediment yields decreased, becoming more comparable to those estimated for the pre-burn period. Total suspended yield in 2001was 61 t km2 (Fig. 15.4b) which was 5.25 times higher than the yield predicted from a regression relationship between annual ﬂow and (pre-burn) suspended sediment yield (11.6 t km2). In 2002 and 2003, the estimates were 24.6 and 28 t km2, both of which were 1.7 times higher than that predicted from the regression. This indicates that suspended sediment yields are beginning to return to baseline levels within a few years following the wildﬁre. This observation concurs with others who suggest the erosion impacts are greatest in the ﬁrst year post-ﬁre and steadily decline over the next few years, returning to baseline values within 3–5 years (e.g., Robichaud et al., 2000; Moody and Martin, 2001).

5.

Discussion and conclusions

In this paper, data from two watersheds draining the Gros Ventre Mountain range of western Wyoming, both with extensive numbers of observations on sediment transport, were explored. At the onset, the two sites were thought to be similar in terms of sediment supply and the relative contributions of bedload, suspended load, and dissolved load to total sediment yield. However, there were apparent differences in sediment loads exported from the two watersheds that imply spatial variability in processes across this province of complex and varied terrain. Our estimates indicate that about 40% more sediment per unit watershed area was exported from Little Granite Creek relative to Cache Creek. Most of this difference was in the suspended sediment and bedload portions where three times more material was estimated to have been exported over the 10-year period of common record. Tentatively, these patterns are linked to differences in rates and modes of hillslope mass wasting that deliver more coarse sediment to Little Granite Creek. An assessment of a landslide database and ﬁeld reconnaissance indicated that large, deep-seated landslides in the Little Granite watershed impinge on the channel and are chronic sources of sediment. Moreover, raveling deposits of glacial and landslide debris in the upper watershed contribute to the suspended sediment load during periods of snowmelt and high intensity rainfall. By comparison, landslides in the Cache Creek watershed are fewer in number and appear less connected to the stream network. Hence, differences in geologic controls in an area of complex terrain may account for the observed dissimilarity in suspended load and bedload; this is an area of on-going investigation in these watersheds. Similar to other studies of sediment transport in gravel-bed streams, measured rates of bedload and suspended sediment in this study were quite variable for a given discharge. High variability has been ascribed to annual and seasonal difference in sediment supply, often in connection with large individual storms, rapid snowmelt, or mass wasting events (e.g., Alvera and Garcia-Ruiz, 2000; Lenzi et al., 2003; Richards and Moore, 2003). However, we were unable to distinguish between-year differences in the datasets from Little Granite or Cache Creeks; measurements for a given year typically fell within the range of those measured for all other years taken together. The inability

404

S. E. Ryan, M. K. Dixon

to distinguish between years could be due to the infrequency with which the highly variable data were collected. However, there were also no documented large-scale disturbances or activities in either watershed during the common period that would account for a sustained, measurable spike in the sediment loads. Hence, the betweenyear similarity is likely valid. Seasonal variability was evident in the suspended sediment records at Little Granite Creek when more material was transported on the rising limb of the snowmelt hydrograph, justifying the use of two rating curves. Similar patterns may be occurring in the bedload component, but measurements were made too infrequently to identify a such an effect. In general, data on the temporal aspects of bedload transport are particularly difﬁcult to obtain, outside of a few instrumented watersheds (e.g., Bunte, 1996; Lenzi et al., 1999; Laronne et al., 2003). Consequently, much of the temporal variability in bedload observations remains unexplained, though newer technologies may provide insight into these processes in the future (Ryan et al., 2005a). Substantial increases in sediment load are often manifest following large-scale disturbances in the watershed. In August 2000, wildﬁre burned portions of the Little Granite Creek watershed, presenting an opportunity to quantify increases in sediment loads associated with large-scale disturbance of forest cover. The primary impact on sediment load was elevated rates of suspended sediment transport observed on the rising limb during snowmelt and during high intensity, late-summer thunderstorms; rates measured during other ﬂow periods were comparable to pre-burn values. The high rainfall intensities generated mudﬂows from shallow soil slips in severely burned areas (similar to that described by Wells (1987)). These were ﬁne-grained, organic rich ﬂows that transitioned into hyperconcentrated ﬂows (based on measured sediment concentrations) as they moved from burned areas into the channel network. With time, the suspended sediment load appears to be returning to baseline values as vegetation re-growth in the watershed stabilizes burned areas, reducing the erosion of sediment from these surfaces. In contrast, there was no quantiﬁable increase in bedload associated with disturbance by wildﬁre. This is attributed to several related factors, driven by the ﬁnding that large-scale, coarse-grained debris ﬂows were not generated in the burned area during the high-intensity storms. Hence, there may not have been a lot of new coarse sediment introduced to the system. Moreover, if substantial coarse sediment had been introduced in the upper part of the watershed, much of it may remain in storage or be exported out of the watershed slowly over time and so the increased transport signal may not show up in the record for decades. Even more problematic, given the inherent variability of the measured rates of bedload transport and the substantial number of samples needed to detect small differences, the effects of largescale disturbance may not be readily detected in the bedload record. Hence, methods other than comparisons of bedload transport data may be needed to quantify the impacts of large-scale watershed disturbances on these geomorphic processes.

References Alvera, B., Garcia-Ruiz, J.M., 2000. Variability of sediment yield from a high mountain catchment, central Spanish Pyrenees. Arctic Antarctic Alpine Res. 32 (4), 478–484.

Spatial and temporal variability in stream sediment loads

405

Buchanan, T.S., Somers, W.P., 1969. Discharge measurements at gaging stations. US Geological Survey, Techniques of Water-Resources Investigations, Book 3, Chapter A8. 65pp. Bunte, K., 1996. Analyses of the temporal variation of coarse bedload transport and its grain size distribution (Squaw Creek, Montana): English translation of Ph.D. dissertation submitted to the Freie Universita¨t Berlin, Germany. USDA Forest Service, Rocky Mountain Forest and Range Experiment Station, General Technical Report RM-GTR-288, 123pp. Clark, M.L., Eddy-Miller, C.A., Mast, M.A., 2000. Cache Creek near Jackson, WY. Environmental characteristics and water-quality of Hydrologic Benchmark Network stations in the West-Central United States. US Geological Survey Circular 1173-C. 115pp. Cobb, E.D., Biesecker, J.E., 1971. The National Hydrologic Bench Mark Network. US Geological Survey Circular 460-D. 38pp. Costa, J.E., 1984. Physical geomorphology of debris ﬂows. In: Costa, J.E. and Fleisher, P.J. (Eds), Development and Applications of Geomorphology. Springer-Verlag, Berlin, Germany, pp. 268–317. Dietrich, W.E., Dunne, T., 1978. Sediment budget for a small catchment in mountainous terrain. Z. Geomorphol. 29, 191–206. Dwire, K.A., Ryan, S.E., Shirley, L.J., et al., 2006. Inﬂuence of herbivory on regrowth of riparian shrubs following a wildland ﬁre. J. Am. Water Res. Assoc. 42 (1), 201–212. Edwards, T.K., Glysson, G.D., 1998. Field methods for measurement of ﬂuvial sediment. US Geological Survey, Techniques of Water-Resources Investigations, Book 3, chapter C2. 80pp. Emmett, W.W., 1980. A ﬁeld calibration of the sediment-trapping characteristics of the Helley–Smith bedload sampler. US Geological Survey Professional Paper 1139. 44pp. Federal Interagency Sedimentation Project. 2005. Federal Interagency Sedimentation Project home page and catalogue. Available on line, accessed on May 16, 2005 at http://ﬁsp.wes.army.mil/ Folk, R.L., 1968. Petrology of Sedimentary Rocks. Hemphill, Austin, TX, 170pp. Gomez, B., Naff, R.L., Hubbell, D.W., 1989. Temporal variations in bedload transport rates associated with the migration of bedforms. Earth Surf. Process. Landf. 14, 135–156. Helley, E.J., Smith, W., 1971. Development and calibration of a pressure-difference bedload sampler. US Geological Survey Open-File Report. 18pp. Laronne, J.B., Alexandrov, Y., Bergman, N., et al., 2003. The continuous monitoring of bedload ﬂux in various ﬂuvial systems. In: Bogen, J., Fergus, T., and Walling D. (Eds), Erosion and Sediment Transport Measurement in Rivers: Technological and Methodological Advances. IAHS-Publ. No. 283. pp. 134–145. Lenzi, M.A., D’Agostino, V.A., Billi, P., 1999. Bedload transport in the instrumented catchment of the Rio Cordon Part I: Analysis of bedload records, conditions, and thresholds of bedload entrainment. Catena 36, 171–190. Lenzi, M.A., Mao, L., Comiti, F., 2003. Interannual variation of suspended sediment load and sediment yield in an alpine catchment. J. Hydrol. Sci. 48 (6), 899–915. Leopold, L.B., 1994. A View of the River. Cambridge, Harvard University Press, 298pp. Leopold, L.B., Emmett, W.W., 1997. Bedload and river hydraulics – inferences from the East Fork River, Wyoming. US Geological Survey Professional 1583. 52pp. Linsley, R.K., Kohler, M.A., Paulhus, H.L.H., 1975. Hydrology for Engineers. 2nd ed. McGraw-Hill, NY. Love, J.D., Christiansen, A.C., 1985. Geological Map of Wyoming. US Geological Survey Special Geologic Map, scale 1:500,000. Montgomery, D.R., Bufﬁngton, J.M., 1997. Channel-reach morphology in mountain drainage basins. Geol. Soc. Am. Bull. 109 (5), 596–611. Moody, J.A., Martin, D.A., 2001. Hydrologic and sedimentologic response of two burned watersheds in Colorado. US Geological Survey Water Resources Investigation Report 01–4122. 142pp. Nolan, K.M., Shields, R.R., 2000. Measurement of stream discharge by wading. US Geological Survey, Water Resources Investigation Report 00–4036. [On CD-ROM]. Parker, G., 1978a. Self-formed straight rivers with equilibrium banks and mobile bed Part 1: The sand-silt river. J. Fluid Mech. 89, 109–125. Parker, G., 1978b. Self-formed straight rivers with equilibrium banks and mobile bed Part 2: The gravel river. J. Fluid Mech. 89, 127–146.

406

S. E. Ryan, M. K. Dixon

Reid, I., Frostick, L.E., Layman, J.T., 1985. The incidence and nature of bedload transport during ﬂood ﬂows in coarse-grained alluvial channels. Earth Surf. Process. Landf. 10, 33–44. Richards, G., Moore, R.D., 2003. Suspended sediment dynamics in a steep, glacier-fed mountain stream, Place Creek, Canada. Hydrol. Process. 17, 1733–1753. Rickenman, D., 1991. Hyperconcentrated ﬂow and sediment transport at steep slopes. J. Hydraul. Eng. 117 (11), 1419–1439. Robichaud, P.R., Beyers, J.L., Neary, D.G., 2000. Evaluating the Effectiveness of Postﬁre Rehabilitation Treatments. General Technical Report RMRS-GTR-63. Ogden, UT. USDA Department of Agriculture, Forest Service, Rocky Mountain Research Station. 85pp. Ryan, S.E., Bunte, K., Potyondy, J.P., 2005a. Breakout Session 2: Bedload Transport Measurement, Data Uncertainty, and New Technologies. Proceedings of the Federal Interagency Sediment Monitoring Instrument and Analysis Research Workshop, September 9–11, 2003, Flagstaff, AZ. US Geological Survey Circular 1276. pp. 16–28. Ryan, S.E., Dixon, M.K., 2002. Bedload movement in a gravel-bed stream. Stream Systems Technology Center, Fort Collins, CO. [On CD-ROM]. Also on-line, accessed on July 21, 2004 at: http://www. stream.fs.fed.us/publications/videos.html Ryan, S.E., Dixon, M.K., Dwire, K.A., 2006. The use of turbidity sensors for monitoring sediment loads following wildﬁre. Proceedings of the 8th Interagency Sedimentation Conference and 3rd Interagency Hydrologic Modeling Conference, April 2–6, Reno, NV. [Available on CD-ROM]. Ryan, S.E., Dixon, M.K., Dwire, K.A., Emmett, W.W., 2003. Historical and on-going hydrologic and sediment transport research at Little Granite Creek near Bondurant, Wyoming. In: Proceedings of the First Interagency Conference on Research in the Watersheds, October 28–30, 2003, Benson, AZ. Agricultural Research Service. pp. 657–662. Ryan, S.E., Emmett, W.W., 2002. The nature of ﬂow and sediment movement at Little Granite Creek near Bondurant, Wyoming. General Technical Report RMRS-GTR-90. Ogden, UT. USDA Department of Agriculture, Forest Service, Rocky Mountain Research Station. 48pp. Ryan, S.E., Porth, L.S., 1999. A ﬁeld comparison of three pressure-difference bedload samplers. Geomorphology 30, 307–322. Ryan, S.E., Porth, L.S., Troendle, C.A., 2005b. Coarse sediment transport in mountain streams in Colorado and Wyoming, USA. Earth Surf. Process. Landf. 30, 269–288. Stednick, J.D., 1991. Wildland Water Quality Sampling and Analysis. Academic Press, New York, 217pp. Troendle, C.A., Nankervis, J.M., Ryan, S.E., 1996. Sediment Transport from Small, Steep-gradient Watersheds in Colorado and Wyoming. In: Sedimentation Technologies for Management of Natural Resources in the 21st Century, Sixth Federal Interagency Sedimentation Conference, March 10–14, 1996, Las Vegas, NV. pp. IX-39 to IX-45. US Geological Survey. 2004. Water quality samples for Wyoming. Department of the Interior, US Geological Survey, Water Resources of Wyoming. Available on-line, accessed on December 12, 2004 at: http://nwis.waterdata.usgs.gov/wy/nwis/qwdata?search_criteria=station_nm&submitted_form= introduction US Interagency Advisory Committee on Water Data, 1982. Guidelines for determining ﬂood ﬂow frequency. Bulletin 17B of the Hydrology Subcommittee. US Geological Survey, Ofﬁce of Water Data Coordination, Reston, Virginia. 183pp. Varnes D., 1978. Slope movement types and processes. In: Landslides Analysis and Control. Special Report 176, National Academy of Sciences, Washington, DC. pp. 11–33. Walling, D.E., 1988. Erosion and sediment yield research – some recent perspectives. J. Hydrol. 100, 113–141. Walling, D.E., Webb, B.W., 1982. Sediment availability and the prediction of storm-period sediment yields. In: Recent Developments in the Explanation and Prediction of Erosion and Sediment Yield. IAHS Publication. 137, 327–337. Waters, T.F., 1995. Sediment in Streams: Sources, Biological Effects, and Control. American Fisheries Society Monograph 7, Bethesda, MD. 251pp. Wells, W.G., 1987. The effects of ﬁre on the generation of debris ﬂows in southern California. In: Costa, J.E. and Wieczorek, G.F. (Eds), Debris Flows/Avalanches. Reviews Engineering Geology 7. Geological Society of America, Boulder, CO, pp. 105–114.

Spatial and temporal variability in stream sediment loads

407

Western Regional Climate Center. 2003. Wyoming snotel sites. Available on line, accessed on April, 15, 2003 at: http://www.wrcc.dri.edu/snotel.html Wohl, E.E., 2000. Mountain Rivers. Water Resources Monograph 14. American Geophysical Union, Washington, DC, 320pp. Wyoming State Geological Survey and the Water Resources Data System. 2001. Preliminary Landslide Map, Wyoming State Geological Survey, Laramie, Wyoming. Available on-line, accessed July 21, 2004 at: http://www.wrds.uwyo.edu/wrds/wsgs/hazards/landslides/landslides.html

Gravel-Bed Rivers VI: From Process Understanding to River Restoration H. Habersack, H. Pie´gay, M. Rinaldi, Editors r 2008 Elsevier B.V. All rights reserved.

409

16 Sediment organisation along the upper Hunter River, Australia: a multivariate statistical approach Joanna Hoyle, Gary Brierley, Andrew Brooks and Kirstie Fryirs

Abstract Freely adjusting gravel-bed rivers are subject to recurrent input of material and frequent, but irregular, reworking of bedload materials. In contrast, many Australian rivers are characterised by low-bedload input and infrequent mobilisation of bedload materials. In this study, the inﬂuence of these boundary conditions upon surface bedmaterial organisation are analysed through application of extensive ﬁeld sampling and multivariate statistics based on a range of bed-material parameters along an 8 km reach of the Upper Hunter River in New South Wales. The Hunter River is a mixedload, gravel-bed river that has been subjected to signiﬁcant disturbance since European settlement. Since European settlement, the channel in the study reach has been transformed from a passive meandering river to a conﬁguration in which the low-ﬂow channel locally adjusts around various bar and bench features. Channel capacity has enlarged locally by a factor of 3, channel alignment has changed via cut-offs (both natural and human-induced), and channel–ﬂoodplain relationships have been altered. The multivariate statistical analyses performed in this study provided a very useful tool for identifying surface facies and the way in which sediment is organised. Bedload surface facies do not demonstrate an equivalent degree of organisation to that documented for fully self-adjusting rivers. A conceptual model is developed that relates the variability and spatial arrangement of differing classes of bed material to the type of geomorphic unit, elevation above the low-ﬂow channel and ease of reworking (i.e., frequency of potential mobilisation). Spatial variability in the sediment mix of the study reach, appraised in light of long-term changes to sediment ﬂux, prompts the need for reﬁnement of bedload transport models for this kind of river. 1.

Introduction

Gravel-bed rivers exhibit complex, yet systematic, patterns of sediment sorting at various scales (Powell, 1998). At the channel-length scale, sorting is exhibited by the E-mail address: [email protected] (J. Hoyle) ISSN: 0928-2025

DOI: 10.1016/S0928-2025(07)11135-4

410

J. Hoyle et al.

downstream reduction in bed surface grain size, despite discontinuities that arise from tributary inputs and valley side sources (Knighton, 1980; Dawson, 1988; Rice, 1994; Pizzuto, 1995). At the reach scale, patterns of sediment sorting are associated with the spatial organisation of pool-bar units and the ﬂow structures that develop around these units. The characteristic bed topographies and sedimentologies of different channel planforms result from, and in turn reinforce, complex ﬂow, boundary shear stress and sediment transport ﬁelds in which different grain sizes adopt different trajectories. Patches with variable sediment mix result from processes of sediment sorting controlled by mesoscale variations in bed topography (i.e., the character and distribution of geomorphic units), whereby induced ﬂow paths route different size fractions along different transport pathways (Parker and Andrews, 1985; Paola, 1989). Particle trajectories are controlled by near-bed ﬂuid vectors and cross-channel and downstream bed slopes (Ikeda and Parker, 1989; Ashworth et al., 1996). The relative effects of gravity and drag force ensure that coarser particles are deﬂected downslope more directly than ﬁner particles for the same near-bed velocity. This process, called ‘topographic sorting’ (Paola, 1989), results in coarser grains becoming concentrated on topographically low surfaces, such as the downstream margins of alternate bars (Lisle et al., 1991), the base of point bars (Bluck, 1971) and scour holes downstream of conﬂuences (Ashworth et al., 1992). Hence, bed-material organisation reﬂects the distribution of geomorphic units over bar surfaces (e.g., upper vs. lower point bar deposits, chute channels, ramp features, etc. (Brierley, 1989). Sediment sorting at the grain scale results in ‘bed structures’, ‘geometric structures’ such as pebble clusters and imbrication or ‘textural structures’ such as armouring and particle interlocking (Wolcott, 1989). Particle projection, exposure, packing and pivoting angle of clasts exert a signiﬁcant inﬂuence on entrainment (Parker and Klingeman, 1982; Wiberg and Smith, 1987; Komar and Li, 1988). Sand shadows may develop in the lee of large obstacle clasts or outcrops of vegetation (Sambrook Smith and Ferguson, 1996). In some instances, bed shear stress at different stages of an event can result in complex patterns and degrees of sediment sorting. In the case of ephemeral streams, infrequent and short duration events may result in a distinct lack of sorting (Laronne et al., 1994). Bed-material organisation along gravel-bed rivers reﬂects the extent to which the channel is freely adjusting. In simpliﬁed terms, gravel-bed rivers can be viewed as a spectrum of forms along an energy gradient, remembering that gravel-bed rivers, by deﬁnition, sit at the higher end of the range of energy conditions under which rivers operate. Along this spectrum, there is a progressive increase in the extent of ﬂoodplain formation, indicating clearer separation of channel and overbank processes. As a consequence, there is a greater degree of textural segregation of deposits in channel and ﬂoodplain zones. Many studies have characterised scales and patterns of gravel organisation along braided rivers. Down-bar ﬁning trends upon mid-channel bars and grain size variation on surfaces of differing elevation across the braidplain are superimposed upon reachscale downstream ﬁning trends (Williams and Rust, 1969; Ashworth and Ferguson, 1989; Best and Bristow, 1993). Patchy gravel organisation at the reach scale results from episodic bedload transport (Ashmore and Parker, 1983; Ashworth and Ferguson, 1986; Ferguson et al., 1992). Complex ﬂow dynamics at zones of ﬂow convergence and

Sediment organisation along the upper Hunter River, Australia

411

divergence, associated with conﬂuences and difﬂuences, vary with the discharge and momentum of the combining and combined ﬂows, junction angle and the depth and orientation of scour zone (Powell, 1998). Along wandering gravel-bed rivers, the reduced number of channels and bars and the greater propensity for lateral channel adjustment relative to braid accretion results in more pronounced down-bar sorting of bed material (Desloges and Church, 1989; Gilvear et al., 2000; Burge and LaPointe, 2005). Along meandering gravel-bed rivers, ﬂow structures that are induced by forces arising from channel curvature, changes in curvature and gradients in bed topography result in strong spatial variations in boundary shear stress, sediment transport and resulting patterns of bed-material size (Dietrich and Smith, 1983; Smith and McLean, 1984; Dietrich and Whiting, 1989). Lisle et al. (2000) noted that adjustments between local boundary shear stress and surface particle size vary for channels characterised by longitudinal, point or alternate (lateral) bars. Around-the-bend patterns of bedmaterial organisation on point bars reﬂect the distribution of ﬂow separation cells associated with the radius of curvature of the bend, resulting in the deposition of ﬁne-grained suspended load materials at the downstream end of point bars in many instances (Jackson, 1976). Under higher energy conditions, point bars may be reworked by ﬂood events that are aligned down-bar, rather than around-the-bend, modifying patterns of bed-material organisation via scour and deposition in chute channels with associated ramp features (McGowen and Garner, 1970; Jackson, 1976; Gustavson, 1978; Bluck, 1987; Blum and Salvatore Valastro, 1989). These various studies have characterised patterns and scales of gravel organisation along fully self-adjusting channels in which input of material is sustained over time and bedload materials are frequently (if irregularly) reworked. Constrictions, whether imposed by valley conﬁnement, cohesive banks, vegetation, bed control structures or other forms, affect channel dimensions, bend curvature, bed topography and gradient, thereby inﬂuencing patterns of erosion and deposition along a reach. These factors are key determinants of bedload ﬂux and the frequency with which bed materials are reworked under given ﬂow conditions. Ultimately, the rate of bedload transport reﬂects the availability of materials of a given size (and ease of entrainment), instream roughness and the frequency of events that are able to mobilise sediments. These relationships, in turn, determine patterns of bed-material organisation along a reach. While our understanding of these considerations is relatively advanced for freely adjusting rivers, few studies have examined equivalent relationships along gravel-bed rivers in alternative settings. For example, a different set of boundary conditions is encountered along mixed-load gravel-bed rivers in tectonically inactive settings where rivers may be subjected to variable (but generally low) rates of bedload input and infrequent rates of reworking. In global terms, the low relief, tectonically stable Australian landscape has relatively few gravel-bed rivers. Existing examples typically reﬂect antecedent conditions, whereby contemporary channels may ﬂow atop (or within) older, cemented, stable deposits, including extensive terrace and ancient ﬂoodplain materials that were deposited under past higher energy conditions (Finlayson and McMahon, 1988; Nanson et al., 1992; Cohen, 2003). Low rates of sediment reworking have been accompanied by low rates of supply in this unglaciated landscape. Bedload sediment supply from upstream is often constrained in these low relief, disconnected landscapes

412

J. Hoyle et al.

(Fryirs et al., in press). In the past, low rates of channel adjustments on valley ﬂoors have promoted the dominance of vertical accretion mechanisms on ﬂoodplains (Nanson and Young, 1981; Nanson, 1986; Nanson and Croke, 1992; Cohen, 2003). Resulting channel capacities are relatively small for these underﬁt streams. Pronounced discharge variability results in infrequent but extreme ﬂoods (Finlayson and McMahon, 1988). However, the geomorphic effectiveness of these events has been limited by small channel capacities, instream roughness and the ease with which ﬂoodwaters spread across valley ﬂoors, dissipating their energy (Brooks and Brierley, 2002). Low slope conditions, along with the infrequent occurrence of bedload transporting events, ensured that the duration of phases under which materials were able to be transported (i.e., periods of effective discharge) were extremely limited (Costa and O’Connor, 1995). As a consequence, it can be inferred that bed materials were infrequently reworked. Along many Australian rivers, these relationships have been comprehensively modiﬁed in response to impacts of human disturbance since European settlement. Incision, expansion and straightening of channels have massively enhanced channel capacity (Nanson and Erskine, 1988; Brooks and Brierley, 1997; Page and Carden, 1998; Fryirs and Brierley, 2001). In some instances, clearance of wood and riparian vegetation triggered dramatic changes to channel morphology (Brooks et al., 2003). Marked reduction in instream roughness greatly enhances the geomorphic effectiveness (and bedload transport) of ﬂood events, as ﬂoods of a higher magnitude are retained within more open channels (Brooks and Brierley, 2004). Enlarged channels have increased accommodation space, enhancing prospects for deposition and remobilisation of materials. Sediments that populate the enlarged channel typically comprise benches or inset ﬂoodplains made up of sand-sized materials (Erskine and Livingstone, 1999; Vietz et al., 2004). Despite the sand-sized texture of materials, the retarding inﬂuence of grasses, herbs, shrubs and large woody debris (Gurnell and Midgley, 1994; Thorne and Furbish, 1995; Bufﬁngton and Montgomery, 1997; Tabacchi et al., 2000) and the infrequent occurrence and limited duration of ﬂood events, may result in relatively low rates of bedload ﬂux. This situation contrasts starkly to unvegetated gravel-bed rivers that are subjected to recurrent bedload transporting events. Given the pronounced differences in boundary conditions experienced by freely adjusting gravel-bed rivers relative to disturbed gravel-bed rivers in Australia, signiﬁcant differences in bed-material organisation and sediment ﬂux can be hypothesised. To address these issues, this study documents a methodology to characterise sediment mixes, describes and explains patterns of sediment organisation and discusses implications for rates of bedload ﬂux. These issues are examined along an 8 km reach of the Upper Hunter River in New South Wales. Patterns of gravel organisation are analysed for the ‘visible’ surface gravel fraction – the most readily (or recently) reworked gravels along the Upper Hunter River. The river has a coarse bedload but ﬁne-grained, cohesive banks. Extensive ﬁeld sampling and multivariate statistical techniques are applied to identify surface facies on bars based on a range of bed-material parameters. The variability and spatial arrangement of these differing facies are explained in relation to the type of geomorphic unit, elevation above the channel and ease of reworking (i.e., potential mobilisation during an event of a given magnitude). The character and distribution

Sediment organisation along the upper Hunter River, Australia

413

of these features are placed in context of the history of post-European settlement disturbance to the river, focussing on the pattern and extent of channel adjustment since 1938, and catchment-scale controls on bedload sediment ﬂux.

2.

Regional setting

The study reach, located 5 km south-west of Muswellbrook (Fig. 16.1), is characterised by a meandering gravel-bed river (sinuosity E1.3) with a valley ﬂoor width of 1–2 km, surrounded by rolling hills of around 100 m relief. It drains a catchment area of 4220 km2. The 8 km reach has an average bed gradient of 0.001 m/m. This reach is the focus of the Upper Hunter River Rehabilitation Initiative (UHRRI) (Brierley et al., 2005). Despite its appearance as a freely self-adjusting channel within a relatively wide, laterally unconﬁned valley (Fig. 16.1), the character and behaviour of the Upper Hunter River are constrained by a range of inherited landscape controls. Bedrock locally abuts the channel margins, on both the bed and banks. Various terrace features, some of which are buried, comprise indurated sediments (including bedded gravels), which locally inhibit channel adjustment. These attributes, along with the lack of any features characteristic of a freely meandering channel, indicate that the Upper Hunter River is a low-to-moderate sinuosity, passively meandering gravel-bed river (Richards, 1982). The ﬂoodplain, including the banks of the study reach, comprises 5–8 m of uniform ﬁne-grained materials (ﬁne sand, silt and clay) atop basal gravels. In contrast, the bed, bars and benches of the channel zone comprise noncohesive sand and gravel. The Upper Hunter has a high-suspended load. Fortnightly low-ﬂow turbidity readings in 2005–2006 ranged from 10 to 150 NTU. Analyses of ﬂoodplain sedimentology, archival records, parish maps and aerial photographs document marked spatial variability in the pattern of channel change since European settlement in the 1820s. The contemporary macrochannel (Van Niekerk et al., 1999) comprises a low-ﬂow channel, ranging in depth from 0.5 to 3.5 m and width from 15 to 40 m, inset within a larger (compound) channel structure. Prior to European settlement the macrochannel width was essentially uniform in width; ranging from 75 to 120 m wide. The contemporary macrochannel ranges from 75 to 600 m wide. The localised but extensive expansion liberated vast quantities of ﬂoodplain material as well as increasing space for sediment storage and reworking. Prior to the emplacement of engineering works in the 1960s, the low-ﬂow channel shifted position within the macrochannel in the zones of extensive adjustment, behaving as an active meandering river. The engineering works have trained the channel such that it behaves as a passive meandering river once more, albeit for quite different reasons to those experienced prior to European settlement. There are three main tributaries to the Upper Hunter River between the study reach and Glenbawn Dam approximately 25 km upstream (Fig. 16.1). Glenbawn Dam was completed in 1958 and cuts off approximately 30% of the Upper Hunter catchment, providing a barrier to any bedload supplied from the Upper Hunter trunk stream. Rouchel Brook and tributaries are bedrock controlled along their entire lengths, with fully conﬁned and partly conﬁned rivers dominating (Brierley and Fryirs, 2005).

J. Hoyle et al.

414

Ba n rri on gt p To

Murrurundi

s

ge Pa ive sR

Hu

k oo Br

nt er

r

Kingdon Ponds

rt Da

N

Hunter Catchment

r Rive

Scone

Lake Glenbawn Brook Rouchel

Macrochannel

Aberdeen

Low flow channel Landscape units

0

10km

20

A

Bar

Remnant plateau

Muswellbrook

B

Plateau slopes

C

Up stream end of study reach

Rugged and hilly

A

Undulating plains

B C

Study Reach

0

1 km

Up stream Keys

2 Milk

N

Keys Bengalla Demo Control Goat s Cockt ail Dan’ s

Edinglassie

Location shown in Figure 16.5

Sandwich Edinglassie 2 Moro

Downstream end of study reach

RL (metres) 145

HUNTER RIVER

Whites

SECTION AA

140 135 130

Gravel: Up to 70% gravel as clasts up to 10cm, minor sand, silt, clay.

125

SECTION BB

140 135

Gravelly sand: well-graded medium to coarse sand with up to 30% gravel, minor silt and clay.

130

Sand: medium to coarse-grained sand with minor gravel and silt and/or clay.

120

Silt and clay: overbank deposits.

125

115

130 125 120

Silty or clayey sand: medium-grained sand with a significant silt or clay component, minor gravel

HUNTER RIVER

Gravelly clay: up to 40% gravel in a clay matrix with only minor sand-sized detritus.

HUNTER RIVER

REFERENCE

SECTION CC

115 110

Figure 16.1. Location map showing Hunter catchment location, Upper Hunter River catchment landscape (including locations of three ﬂoodplain cross-sections near the study reach), a study reach map and three cross-sections showing ﬂoodplain sedimentology in the vicinity of the study reach. These cross sections are modiﬁed from Williamson (1958). Please note that these cross-sections have been drawn looking upstream, hence, the left bank in the ﬁgure is actually the right bank in the ﬁeld.

Sediment organisation along the upper Hunter River, Australia

415

Bedrock is exposed on the channel bed and many banks. Hence, Rouchel Brook supplies very little in the way of bedload or suspended load sediment. The Pages River is the most signiﬁcant tributary and has undergone signiﬁcant incision and channel expansion since European arrival to the area. This change to the river has exposed bedrock on the bed and has also lead to the creation of an extensive slug of bedload sediment (Bartley and Rutherfurd, 1999; Fryirs and Brierley, 2001) reaching almost to its conﬂuence with the Hunter. The third main tributary, Dart Brook, is a low slope, low capacity, highly sinuous, entrenched stream with cohesive banks and very little gravel bedload. This very low energy stream has limited capacity to move what little bedload is available (Fryirs et al., in press). This leaves the study reach with a limited long-term supply of bedload from upstream. Bulk sieving at 18 sites throughout the study reach indicated that there is signiﬁcant variability in grain size distribution within bars. Surface D50 ranged from 43.3 to 2.3 mm and subsurface D50 from 26.8 to 6.5 mm. Surface D50 to subsurface D50 ratios ranged from 2.6 to 0.3, giving an indication of the degree of surface coarsening. Signiﬁcant variability in ﬂow regime results in long periods of relative inactivity punctuated by occasional high magnitude-low frequency events. Fig. 16.2 shows a reconstructed 66-year hydrograph for the Muswellbrook bridge gauge (stn no. 210002), which is located approximately 6.2 km upstream of the study reach. Log Pearson calculations were used to determine the following ARI data (based on a 4100 year ﬂood history): 1 year ARI ¼ 11 m3/s, 2 year ARI ¼ 350 m3/s, 5 year ARI ¼ 1000 m3/s, 10 year ARI ¼ 1650 m3/s, 20 year ARI ¼ 2025 m3/s, 100 year ARI ¼ 4944 m3/s. Under the current ﬂow regime, 90% of the time ﬂows in the river are o12 m3/s. Ten percent of the time ﬂows are o1 m3/s. Glenbawn Dam has resulted in increased low ﬂows and decreased peak ﬂows. The 1955 event is the largest ﬂood on record (5685 m3/s). This event resulted in channel expansion and lead to the formation of extensive bars that are essentially in the same place today. Since 1955 there have Upper Hunter River Hydrograph at Muswellbrook

Maximum Discharge (m3/s)

6000

5000

4000

3000

2000

1000

2004

2002

2000

1998

1996

1994

1992

1990

1988

1986

1984

1982

1980

1978

1976

1974

1972

1970

1968

1966

1964

1962

1960

1958

1956

1954

1952

1950

1948

1946

1944

1942

1940

1938

0

Year

Figure 16.2. Reconstructed hydrograph of the Hunter River from 1938 to 2004 based on stage heights taken at the Muswellbrook bridge gauge and Muswellbrook Weir. Expansion of the macrochannel resulted in changes to the river cross-section at the Muswellbrook gauge site between 1928 and 1960. The hydrograph has been reconstructed to correct for these changes.

J. Hoyle et al.

416

been two events greater than a 1:20 year event (1971 and 1992) and 11 events greater than a 1:5 year event. The period post 1955 has been one of aggradation within the study reach, with extensive deposition and reworking of bars occurring in the 1970s.

3.

Field methodology

Understanding of large-scale sediment ﬂuxes requires an appreciation of the types of sediment mix available for transport, their spatial organisation within a channel and the relative mobility of particles in different locales. Accurate representation of the channel topography for analysis of mega and macroforms was gained from airborne laser scanning (ALS). This was acquired in winter 2003 to maximise the beneﬁts of the willow tree defoliation within the riparian zone (Cohen et al., 2004). These data were supplemented and veriﬁed by survey data collected using a total station at three locations (1100–1860 survey points) as well as 17 surveyed cross-sections. As ALS is unable to collect accurate data below the water surface, a total station survey of the topography within the low ﬂow channel was completed to augment the ALS data. This information was stitched into the ALS data to develop a 3 m pixel resolution digital elevation model (DEM) of the study reach. Using the DEM, along with two sets of ortho-rectiﬁed stereo pairs of aerial photographs (ﬂown in 1998 and 2004), the larger scale geomorphic units (megaforms) were accurately mapped using ARCMap GIS software. Eleven main bars were identiﬁed in the 8 km study reach. Within each of these bars, the major geomorphic units (at the macroform and mesoform scale) were identiﬁed, surveyed and mapped. One hundred and three geomorphic units were identiﬁed, comprising 15 different types (see Table 16.1). The boundaries between the various geomorphic units were distinguished on the basis of abrupt breaks in slope. Field maps were tied into the DEM and ortho-rectiﬁed aerial photographs by creating overlays in ARCMap. A quick, reliable and easy method was developed to analyse gravel organisation at the microform scale for differing geomorphic surfaces. This entailed a synthesis of available literature on the scale, grain size attributes, packing arrangement and structural patterns of sediment storage features expected along gravel-bed rivers (Karcz, 1972; Gustavson, 1978; Richards, 1982; Brayshaw et al., 1983; Reid and Frostick, 1984; Bluck, 1987; Wilcock and Southard, 1989; Church et al., 1998; Wittenberg, 2002; Strom et al., 2004). A simple classiﬁcation was established in which a series of variables known to inﬂuence sediment mobility (namely maximum grain size, sorting, support structure, packing, presence of imbrication or other surface structures, degree of censoring, presence of any surface alteration and the degree of vegetation cover) was each split into categories (see Table 16.2). One hundred and sixty-three sample sites were selected covering all geomorphic units identiﬁed within the eleven main bars. Sample sites were approximately 3 by 3 m (at times restricted by the size and shape of the geomorphic unit). Care was taken to ensure sample site locations within each geomorphic unit were as consistent across bars and units as possible. Generally the sites were at the upstream end of each unit. If signiﬁcant downstream ﬁning or coarsening was evident, or several signiﬁcantly different

Sediment organisation along the upper Hunter River, Australia Table 16.1. 2005).

417

Deﬁnition of geomorphic units found in the study reach (Brierley and Fryirs,

Geomorphic unit

Code

Deﬁnition

Bench

1

Ridge

2

Drape

3

Scour holes

4

Ramp

5

Platform

6

Avalanche face

7

Chutes/overﬂows

8

Upstream sheet

9

Apex sheet

10

Inset bar

11

A distinctly stepped feature deposited against the bank of the macrochannel. They may have multiple surfaces reﬂecting different phases of channel contraction and reworking. They are composed largely of interbedded gravels and sands with occasional ﬁner material. They are elongate with a straight to gently curved planform, ﬂanking one or both banks. Major in-channel sediment storage unit, often situated atop bar deposits. May have obliquely attached mud-rich drapes with a convex geometry on relatively steep banks. Depositional unit reﬂecting channel contraction of cut-and-ﬁll processes. Linear, elongate deposits. May be sinuous or irregular, resting atop bar platform, comprising a compound bank-attached bar. May be formed downstream of vegetation or other obstructions on the bar surface. Obliquely attached mud-rich drapes with a convex geometry on relatively steep banks. Depressions or holes formed by the erosive power of moving water. Commonly, but not exclusively, found at the upstream end of bars. Coarse-grained inclined forms commonly occurring on bars where sediment is scoured from the head of a bar and ramped up before being deposited on a platform towards the tail of the bar. These features coarsen downstream. A relatively ﬂat low angle surface occurring on bars where sediment has been deposited at high ﬂow. The gravel becomes progressively ﬁner grained downstream. Slip faces occurring at the downstream end of bars. These features usually exhibit downslope coarsening. Channels that dissect a bar surface. These channels are formed by water at high ﬂow allowing the ﬂow to take a direct short cut across the top of bars or to descend rapidly to a lower level as water level drops in the waning stages of an event. A common feature on compound point bars and islands. Low elevation features occurring at the most upstream area of the bar. These areas are frequently inundated by small events and may form when bedload material is mobilised delivering it from the low ﬂow channel up onto these sheets. They generally have a similar composition to bedload in the low ﬂow channel but with additional ﬁne-grained materials. Relatively ﬂat low angle surfaces that occur around the apex of bars generally at a lower elevation than the remainder of the bar. They may exhibit downstream ﬁning. Low elevation, relatively ﬂat unit bars that sit between the low ﬂow channel and the greater bar. They generally occur around the edge of the downstream half of the greater bar and particularly below avalanche faces and are readily reworked during small events.

J. Hoyle et al.

418 Table 16.1 (continued ) Geomorphic unit

Code

Deﬁnition

Unit Bar

12

Ridge and scour complex

13

Rifﬂe

14

Sand patch

15

Unit bars occur at the megaform scale but are absent of inset smaller scale geomorphic units. They are essentially made up of a rounded platform that may exhibit downstream ﬁning or have drapes of ﬁnes around the waters edge. In order to examine the patterns of sediment organisation on these bars and therefore compare them with other bars, they are included at the geomorphic unit scale as a unit in their own right. These are areas of undulating topography typically occurring at the downstream end of bars. These areas could be split into individual smaller geomorphic units (ridge or scour) but, due to their small size, for the purposes of this study they were clumped. Topographic highs along an undulating longitudinal proﬁle. They occur at characteristic locations, typically between bends (the inﬂection point) in sinuous alluvial channels. Patches of ﬁne sediments generally deposited on top of other geomorphic units.

sediment mixes were found within a geomorphic unit, additional samples were taken. Surface sediments in the low-ﬂow channel were excluded from this analysis because the methods used on the bars could not be applied in this permanently wetted environment and turbidity levels prohibited a visual analysis. At each site the maximum clast size (a measure of the ﬂow required to mobilise the entire sediment mix) and range of sediment sizes (based on the Wentworth scale of boulder, cobble, pebble, granule and sand) was determined via the measurement of clast b-axis. On the larger geomorphic units, systematic b-axis measurement was carried out along a series of transects established across the unit (the Wolman method) in order to establish whether any downstream ﬁning was present (Wolman, 1954; Leopold, 1970). In these cases a minimum of 100 clasts were collected. A visual guide to geometric and textural structures was used to determine the presence of any imbrication or other surface structures on the given gravel surface, as well as the degree of sorting and surface censoring (see Fig. 16.3). Any coarse surface layer lacking in ﬁne-grained materials is termed censored (Carling and Reader, 1982), while the presence of ﬁne-grained materials that have settled into gravel interstices is termed embedded. Surfaces neither censored nor embedded are termed non-censored. Packing was measured as the ease of dislodging clasts from a surface. If clasts overlap such that they are held ﬁrmly in place and cannot be dislodged easily, then the bed of the river is tightly packed. For instance, older exhumed gravel lag layers are often densely packed. If there is no overlap between clasts, and they can be easily dislodged, the bed is considered to be loose. In mixed load rivers such as the Hunter, a surface may contain a wide range of sediment sizes. Clasts of similar size may not overlap and the voids between them may comprise smaller sized grains creating a

Sediment organisation along the upper Hunter River, Australia Table 16.2.

419

Methods and codes used to classify surfaces at each sample site.

Variables tested

Methodology

Category code

Maximum clast size

Largest clast present at the sample site based on the Wentworth scale: boulder, cobble, pebble, granule, sand. Classiﬁed using Fig. 16.3 as a guide: well, moderate, poor. Very packed – cannot be easily kicked apart. Packed – can be kicked apart with minimal effort. Moderately packed – very easily kicked apart or can be broken up with hand. Loose – moderately loose and can be picked up by hand. Very loose – loose and can be picked up by hand. If well sorted sands then they are deﬁned as massive. Otherwise, a number of clasts are removed and if others fall into their place or those removed sit directly above other clasts then they are considered clast supported. Clasts surrounded in ﬁne-grained materials are considered matrix supported. Buried – no coarse clasts visible. Embedded – interstices are ﬁlled with ﬁne-grained materials. Censored – top layer of sediment (1 clast thick) is devoid of ﬁne-grained materials. Non-censored – coarse and ﬁne clasts are mixed with no obvious surface layer. Dense – 95–100% vegetation cover. Vegetated – 75–95% vegetation cover. Patchy – 25–75% vegetation cover. Sparse – 5–25% vegetation cover. Unvegetated – 0–5% vegetation cover. Observational based on knowledge of descriptions in literature (Richards and Clifford, 1991). Observational based on knowledge of descriptions in literature (Laronne and Carson, 1976; Brayshaw et al., 1983; Bluck, 1987; Wolcott, 1989; Hassan and Reid, 1990; Wittenberg, 2002; Brierley and Fryirs, 2005). Type of structure noted. Surfaces with a packing density of 1 were examined to determine whether they were ancient lithoﬁed structures (buried terraces), whether they were immobile due to baked hard interstitial ﬁne-grained materials or neither.

Boulder – 1, cobble – 2, pebble – 3, granule – 4 or sand – 5

Degree of sorting Degree of packing

Support structure

Surface ﬁnes

Vegetation

Imbrication Clast clusters (or other surface structure)

Baking (surface alteration)

Well – 1, moderate – 2 or poor – 3 Very packed – 1, packed – 2, moderately packed – 3, loose – 4 or very loose – 5

Clast – 1, matrix – 2 or massive – 3

Buried – 1, embedded – 2 censored – 3 or non-censored – 4

Dense – 1, vegetated – 2, patchy – 3, sparse – 4 or unvegetated – 5

Present -1 or absent – 2 Present – 1 or absent – 2

Present – 1 or absent – 2

J. Hoyle et al.

420 Table 16.2 (continued ) Variables tested

Methodology

Bar

Note the bar from which the sample Numbered 1 – 11 from upstream to was taken. downstream Note the geomorphic unit from which Coded as in Table 16.1. the sample was taken. Based on ﬁeld survey and GIS U/s and low – 1, U/s and mid – 2, U/s mapping. Bar split into a 3 3 and high – 3, mid and low – 4, grid, where longitudinal and mid and mid – 5, mid and high – 6, vertical locations were each split D/s and low – 7, D/s and mid – 8 or into thirds. D/s and high – 9

Unit Location on Bar

Category code

dense surface where clasts are not easily dislodged. Drying of cohesive ﬁne-grained materials results in embedded surfaces becoming cemented and extremely hard. These surfaces may also be considered packed. When subjected to prolonged wetting, these surfaces may be remobilised relatively easily. Although this investigation focuses on the gravel surface, the kick-method used to determine packing (Table 16.2), also exposes the subsurface enabling the presence of any armouring and the type of support structure (clast support, matrix support or massive) to be noted. Other measures used to characterise the sediment mix included the extent of vegetation coverage and the presence of any surface alteration (generally baking – noted by the presence of a ‘crust’).

4.

Data analysis methodology

The classiﬁcation system used to analyse sediment mix resulted in a series of categorical data for a number of different variables. An appraisal of patterns of statistically discrete forms of sediment organisation on different surfaces and how these forms of organisation relate to their location was performed using multivariate statistical analysis. The software packages used were PAST (Ryan et al., 1995) and ADE-4 (Thioulouse et al., 1997). Multivariate methods base their comparisons of two or more samples on the extent to which samples share particular attributes. The two multivariate methods used in this investigation were hierarchical clustering and multiple correspondence analysis (MCA). Hierarchical clustering is founded on similarity coefﬁcients calculated between every pair of samples. It generates a ‘dendrogram’ which clearly shows clusters of similar samples, displaying the statistical similarity of the samples in each cluster (Clarke and Warwick, 2001). This method was used to indicate whether there are common forms of sediment organisation within the study reach. The descriptive sediment data were converted to binary format and the Jaccard similarity measure was used. Jaccard similarity is a single, simple measure that is widely used for comparing biological communities where the measure reﬂects the list of species present at each site but not their abundance (Van Sickle, 1997). In this study, the measure represents the proportion of the total number of sediment descriptive categories that

Sediment organisation along the upper Hunter River, Australia

Well Sorted

Moderately Sorted

421

Poorly Sorted

Figure 16.3. Characteristic gravel structures characterised in the literature (including imbrication, pebbles clusters, transverse ribs/clast dams, censoring, embedding/inﬁll, interlocked/packing, dumped/non-censoring).

422

J. Hoyle et al.

are shared by both sites and ranges between 0 (no descriptive variables in common) and 1 (identical sediment description). The Jaccard similarity between sites i and j, with sij ¼ sji is given by sij ¼ Cij/(Cij+Ui+Uj), where Cij is the number of common descriptive variables at the two sites, and Ui, Uj are the numbers unique to each site (Digby and Kempton, 1987). PAST was used to carry out this similarity and hierarchical cluster analysis, from which a dendrogram (or cluster tree) was plotted. Dendrograms work by combining samples, progressively clumping clusters with the highest group average similarity. The algorithm used to generate the dendrogram in this study was the unweighted pair-group average (UPGMA). This algorithm joins clusters based on the average distance between all members in the two groups. Location-related patterns at different scales were assessed using ADE-4 to run MCA. MCA is a technique for displaying the rows and columns of a sample by variable data matrix with categorical variables as points in dual low-dimensional vector space using the w2 distance function (Greenacre, 1984; Benzecri, 1992). Each sample site is represented by a point on the MCA scatter plot and its proximity to other points is a representation of its similarity based on the descriptive variables. A cloud of points with the same location at a particular scale may be linked to their barycentre (centre of gravity), which may be considered as the mean of the cloud of points. By visualising the relative proximity of the barycentres we get an understanding of how similar different locations are in terms of their sediment organisation. MCA also allows us to plot the modalities of each descriptive variable for a particular scatter plot (Clement and Pie´gay, 2003). This allows us to see which descriptive variables deﬁne the key differences between the barycentres and therefore how sediment organisation varies between locations. Three between-class MCA analyses were run. The ﬁrst aimed to maximise the variance between bars to see if sediment on particular bars could be described differently to those on other bars. The second analysis aimed to maximise the variance between locations within a bar. Bars were split into thirds longitudinally and vertically (corresponding with lateral) resulting in a 3 3 grid on each bar. Each of these positions was numbered resulting in 9 class variables, as outlined in Table 16.2. The aim of this was to see if particular longitudinal or lateral positions on a bar tended to have particular forms of sediment organisation. The third analysis aimed to maximise the variance between geomorphic units to see if particular geomorphic units were associated with particular forms of sediment organisation. Randomisation tests (Manly, 1991) were run for each analysis using 10,000 permutations. All showed very good differentiation between sites at each scale, particularly with bars. An analysis of historic aerial photographs (stereo pairs from 1938, 1955, 1969, 1979, 1989, 1998 and 2004) and creation of overlays of sampling site locations in ARCMap was performed to estimate the earliest year that key geomorphic units on the study bars were created. One dimensional hydraulic modelling was then carried out using HEC-RAS 3.2.1 (U.S.A.C.E., 2004) to determine the frequency and depth of inundation for each of these units. ARI values were reconstructed from a 66-year ﬂood history at the Muswellbrook bridge gauge (stn no. 210002) (Fig. 16.2), and 79 cross-sections extracted from the 3 m DEM.

Sediment organisation along the upper Hunter River, Australia 5. 5.1.

423

Results Reach scale

Mapping of the larger scale geomorphic units (megaforms) using ARCMap GIS software showed that under low-ﬂow conditions, approximately 13% of the contemporary macrochannel is made up of bars and islands, benches comprise 67% and the low-ﬂow channel ﬁlls the remaining 20%. The bars in the study reach are generally compound point bars (making up 83.1% of all bars) but there are also compound lateral bars (9.6%), unit lateral bars (4.8%), mid-channel bars (2.1%) and islands (o1%). The randomisation test between the 11 bars showed very good differentiation between bars indicating that there are sediment organisation differences between bars. Fig. 16.4a shows the scatter plot resulting from the between-class MCA conducted on the 11 bars. This plot is in star format, which allows us to see the location of all 163 sample points and also the position of the barycentres of each group. The trajectory graphs shown in Fig. 16.4b indicate the modalities of each of the variables. We can see that the F1 axis (horizontal axis) is structured by grain size and support structure, whereas the F2 axis (vertical axis) is mainly inﬂuenced by packing and vegetation. The two bars furthest downstream in the reach (Sandwich bar – 10 and Whites bar – 11) occur on the far left side of the scatter indicating that they tend to be ﬁner than the other bars. However, an overall downstream ﬁning trend is not apparent over this 8 km reach. The barycentres on this scatter, other than point 8, are fairly evenly spread. This indicates that groupings of similar bars are not evident. Rather, all are equally different, other than Edinglassie bar (bar 8), which is quite different to the other bars. The scatter of the modalities indicate that relative to other bars, sample sites at Edinglassie bar have a higher tendency to be coarse, packed, imbricated and unvegetated.

5.2.

Within-bar and geomorphic unit scale

Similar assemblages of geomorphic units occur in similar patterns for each compound bar, regardless of whether it is a compound point bar or compound lateral bar (Fig. 16.5). Platforms constitute the largest area of the compound bars. They are generally bounded by a ramp at their upstream end, an avalanche face at their downstream end and ridges and benches at their respective proximal and distal lateral edges. Compound point bars tend to have proportionally larger apex sheets and more frequent ridges and scours than compound lateral bars, which tend to have proportionally larger upstream sheets. Rarely observed unit bars do not have inset geomorphic units and generally comprise a downstream ﬁning platform. Fig. 16.4c and d relate sediment organisation to within-bar geomorphic unit. The plot shows that the units with the most clearly different sediment organisation are benches and inset bars. Benches comprise densely vegetated, massive ﬁnes. Inset bars are made up of sparsely vegetated, clast-supported cobbles. The unit named ridge and scour complex, which was clumped for ease of analysis at the data collection stage, plots very closely to ridges, scours, overﬂow channels and drapes.

J. Hoyle et al.

424

1. U/s Keys 2. Keys 3. Bengalla 4. Dan’s 5. Control Goats 6. Control Goats Island 7. Edinglassie 2 8. Edinglassie 9. Moro 10. Sandwich 11. Whites

a)

b)

-

-

c)

1. Bench 2. Ridge 3. Drape 4. Scour 5. Ramp 6. Platform 7. Avalanche Face 8. Overflow/Chute 9. U/s Sheet 10. Apex Sheet 11. Inset Bar 12. Unit Bar 13. Ridge & Scour 14. Riffle 15. Sand Patch

d)

1. U/s – Lower 2. U/s – Mid 3. U/s – Higher 4. Mid – Lower 5. Mid – Mid 6. Mid – Higher 7. D/s – Lower 8. D/s – Mid 9. D/s – Higher -

e)

f)

Figure 16.4. (a) MCA scatter plot showing similarity of sediment organisation between the 11 bars in the study reach. (b) Projection of the modalities indicating which variables describe the differences in sediment organisation between bars. (c) MCA scatter plot showing similarity of sediment organisation between geomorphic units. (d) Projection of the modalities describing how geomorphic units vary. (e) MCA scatter plot showing similarity of sediment organisation at different vertical and horizontal locations on a bar. (f) Projection of the modalities indicating how these sites vary in their organisation.

Sediment organisation along the upper Hunter River, Australia

425

Figure 16.5. Schematic of geomorphic unit locations on bars (based on the 2004 air photograph) and oblique photograph of geomorphic units in a section of the study reach.

This indicates that clumping this unit is unlikely to have affected results. Unit bars and platforms plot closely together, supporting the notion that unit bars effectively comprise a rounded platform. Fig. 16.4e and f display the differences in sediment organisation related to lateral and longitudinal position on bars. This MCA plot shows the most distinct groupings of all the plots, with clear differences in sediment organisation corresponding to vertical/lateral position on the bar. For instance points 3, 6 and 9 represent the highest third of the bar (also the part of the bar most laterally distant from the low-ﬂow channel). The modality plots show us that this part of the bar generally comprises highly vegetated, matrix supported sands and granules. Points 2, 5 and 8 represent the mid-elevations, which are generally made up of poorly sorted, embedded cobbles with sparse vegetation and a high degree of packing and baking. Points 1, 4 and 7 represent

J. Hoyle et al.

426

the lower elevations, close to the low-ﬂow channel. These areas are likely to be loose, unvegetated and censored with presence of imbrication and clustering. Of interest here is that the apex of the bar (mid-section) is likely to be much coarser than the upstream and downstream sections. At no elevation is there a pattern indicating a downstream reduction in sediment size. Overall, these results indicate that sediment organisation varies according to the geomorphic units present on a bar and that some geomorphic units are clearly different in their sedimentary makeup than others, whereas some units have enough variability within them to mask obvious patterns. For instance, platforms are the largest units found on bars and they may constitute a variety of sediment organisation. Therefore, sediment organisation at a particular site appears to depend both on its location within the bar as well as the unit on which it is found.

5.3.

Sediment mix scale

The dendrogram plotted from the hierarchical cluster analysis is shown in Fig. 16.6. This dendrogram clearly shows three distinct groups, each of which has been split into a few smaller groups. The ﬁrst split separates cluster 1 from clusters 2 and 3 and occurs at a similarity of 0.2, meaning that samples in cluster 1 are 80% different to other samples. Cluster 2 separates from cluster 3 at 0.27 similarity and the subgroups of each of the three main clusters separate between 0.3 and 0.4 similarity. The seven distinct groups identiﬁed from this dendrogram are therefore at least 60% different from each other and the forms of sediment organisation represented by each group are considered to be distinct surface facies. The speciﬁc attributes of these seven statistically distinct surface facies are presented in Table 16.3 and an example photograph of each is shown in Fig. 16.6. The facies can generally be described as follows:

Surface facies 1a: Well sorted massive sands. Generally packed and densely vegetated. Surface facies 1b: Well sorted, matrix supported sands and granules. Generally loose and vegetated. Surface facies 2a: Poorly sorted, clast-supported cobbles and boulders. Generally vegetated and embedded in ﬁnes, which have baked hard resulting in a packed surface. Surface facies 2b: Poor to moderately sorted, matrix supported pebbles and cobbles. These appear to have been rapidly deposited and are generally moderately packed and vegetated. Surface facies 3a: Censored cobbles and pebbles. These are clast supported and generally very loose with poor to patchy vegetation. Surface facies 3b: Censored pebbles and granules. These are clast supported and generally very loose with poor to patchy vegetation. Surface facies 3c: Censored and imbricated cobbles and boulders with occasional clusters. Generally packed and free of vegetation.

0.9 0.8 0.7

Similarity

0.6 0.5 0.4 0.3 0.2 0.1

10

20

30

2a

3b

3a

3c

40

50

60

70

80

2b

90

100

110

120

1a

1b

130

140

150

160

Number of samples 3c

3a

3b

2a

2b

1b

1a

Scale:

Sediment organisation along the upper Hunter River, Australia

1.0

500mm

427

Figure 16.6. Dendrogram (cluster tree) indicating the clustering of the sample sites into groups of similar sediment organisation (surface facies) including example photographs of each surface facies.

J. Hoyle et al.

428 5.4.

Relative abundance of surface facies

The most abundant surface facies found within the macrochannel is 1a (massive sands, densely vegetated and packed). This surface facies is estimated to cover 87% of the macrochannel surface, excluding the low-ﬂow channel. This surface facies is predominantly found on benches which make up the largest proportional area of the macrochannel within the study reach. If we focus only on bars, the most abundant surface facies is 3a (36%), comprising moderately well sorted, loose and censored cobbles with patchy vegetation. The relative abundance of each of the surface facies on bars is summarised in Table 16.3.

6.

Reach hydraulics: frequency of inundation and mobilising ﬂows

Output data from HEC-RAS were used to calculate the recurrence with which units were inundated. Wolman sediment size data were used to assess the potential mobility of sediments on each geomorphic unit based on standard sediment transport equations (HR Wallingford, 1990) embedded within a simple one-dimensional model known as Geomorphic Assessor 2.1 (Parfait, 1998). This provides a ﬁrst order approximation of the event magnitude required to mobilise D84 of the sediment on each unit. Results from HEC-RAS and Geomorphic Assessor give an indication of the number of times that each unit has potentially been reworked since they were formed (Table 16.4). However, the actual frequency of reworking of each surface is likely to be much greater (longer) than the values reported, due to the presence of vegetation on most surfaces (Thornes, 1990). Benches are not included in this analysis due to the presence of dense grass that protects these surfaces. The frequency of inundation reﬂects unit elevation. Most bar units are inundated during 1–2 year events whereas benches tend to require a 5–10 year event. Frequency of mobilisation is ﬁrst limited by elevation and then by sediment size at that elevation. Relating these results to the surface facies on each unit enables the degree of sediment organisation to be explained. On average, mobilisation of facies 1a requires a 1:4 year event (due to elevation rather than size), facies 1b is mobilised in a 1:2 year event, facies 2a requires a 1:4 year event, facies 2b a 1:3 year event, facies 3a requires a 45 year event, facies 3b only a 1:2 year event and facies 3a requires a 410 year (bankfull) event.

7. 7.1.

Discussion Use of statistics to analyse sediment organisation and quantify surface facies

The multivariate statistical analyses performed in this study provided a very useful tool for identifying surface facies and the way in which sediment is organised. They allowed a large number of samples with numerous descriptive variables, collected across complex spatial scales, to be easily analysed and for any recurring patterns to be displayed. Seven distinct forms of sediment organisation were identiﬁed within the study reach. The location of these different surface facies relates strongly to the

Attributes of surface facies including locations that they are found and the proportion of bars covered by each.

Description

Unit location

Maximum clast

Sorting

Support structure

Packing density

Imbrication Clusters

Censoring

1a

Massive sands, packed and densely vegetated Matrix supported sands and granules, loose and vegetated Poorly sorted and clast-supported cobbles and boulders, embedded in ﬁnes, baked and vegetated Poorly sorted and matrix-supported cobbles and pebbles, appear rapidly dumped Censored and loose cobbles and pebbles with patchy vegetation Censored and loose pebbles and granules, generally unvegetated Censored and imbricated cobbles and boulders with occasional clusters, unvegetated

Benches, overﬂows, scours Benches, U/s sheets, unit bars

Sand

Well

Massive

Packed

No

No

NonNo censored

Densely vegetated

Granule

Well

Matrix

Very loose

No

No

NonNo censored

Vegetated

7.4

Platforms, overﬂows, ridges, scours

Cobble

Poor

Clast

Very packed

No

No

Embedded

Yes

Vegetated

11.6

Ridges, scours, benches

Cobble

Poor

Matrix

Packed

No

No

NonNo censored

Vegetated

20.2

Platforms, rifﬂes, unit bars, ramps, ridges

Cobble

Mod

Clast

Very loose

No

No

Censored

No

Patchy

35.8

U/s sheets, inset bars, apex sheets

Pebble

Poor

Clast

Very loose

No

No

Censored

No

Unvegetated

3.6

Rifﬂes, apex sheets, U/s sheets, inset bars, ramps

Boulder

Mod

Clast

Packed

Yes

Yes

Censored

No

Unvegetated

4.4

1b

2a

2b

3a

3b

3c

Baking Vegetation

Bar coverage (%) 23.1

429

Surface facies

Sediment organisation along the upper Hunter River, Australia

Table 16.3.

J. Hoyle et al.

430

Table 16.4. Frequency of inundation and reworking of differing geomorphic surfaces, and associated gravel populations. Bar name

US Keys

Gravel units

Platform Overﬂow Ridge US sheet Keys Bridge Platform Scour Apex sheet Inset bar Bengalla Ramp Inset bar US sheet Platform Ridge Avalanche face Dans Platform Inset bar Ridge Sand patch US sheet Control Goats Rifﬂe Apex sheet Ramp Platform Inset bar (Island) Ridge Mid channel bar Avalanche face Edinglassie 2 Apex sheet Ramp Ridge Platform Unit bar Edinglassie Platform US sheet Inset bar 1 Inset bar 2 Avalanche face Moro Platform Apex sheet Sandwich Platform Inset bar Apex sheet Avalanche face

First observed on aerial

Inundation ARI

Mobilisation ARI

Surface facies

1979 1979 1979 1989 1979 1989 1989 2004 1969 2004 1969 1969 1969 1969 1938 2004 1979 2004 2004 2004 1989 1955 1955 2004 1955 2004 1955 1979 1979 1979 1979 1989 1955 1979 2004 1998 1955 1955 1955 1955 2004 1979 1955

1:1 1:2 1:2 1:2 1:2 1:2 1:2 1:2 1:2 1:1 1:2 1:2 1:2 1:2 1:2 1:1 1:2 1:2 1:1 1:1 1:1 1:2 1:2 1:2 1:2 1:1 1:2 1:2 1:2 1:2 1:2 1:2 1:2 1:2 1:1 1:1 1:2 1:2 1:2 1:2 1:2 1:2 1:2

1:3 1:3 1:3 1:2 41:10 1:2 41:10 41:10 41:10 1:2 1:2 1:2 1:2 1:5 41:10 1:3 1:2 1:2 1:2 1:2 1:10 41:10 41:10 1:10 1:2 1:1 1:4 1:4 1:10 1:4 1:4 1:4 41:10 41:10 1:2 41:10 41:10 1:4 1:2 1:5 1:2 1:2 1:3

2a 2a 2a/2b 1b 3a 2a 3a 3c 3a/3c 3a 1b 3a 2b 3a 2a/3a/3b 3b 3a/3b 1b 1b 3a 3a 3a 3a 2a/2b/3c 2a/2b/3a 3b 2b 2a 3a 2a 2a 3a/1b 3c 3c 3c 3c 3c 3a 2a 3a 3a 1b 3a

(Bankfull) (Bankfull) (Bankfull) (Bankfull)

(Bankfull)

(Bankfull) (Bankfull)

(Bankfull) (Bankfull) (Bankfull) (Bankfull)

Sediment organisation along the upper Hunter River, Australia

431

Table 16.4 (continued ) Bar name

Gravel units

First observed on aerial

Inundation ARI

Mobilisation ARI

Surface facies

Whites

Apex sheet Ridge Inset bar US sheet Platform

1955 1955 2004 2004 1955

1:2 1:2 1:2 1:2 1:2

1:2 1:4 1:2 1:5 1:4

3a/3b/2b 3a 3a 3a 1a

geomorphic unit on which they are found and more speciﬁcally to its elevation above the low-ﬂow channel. 7.2.

Conceptual model of surface facies formation

The sediment mixes of bar deposits in the study reach generally range from cobbles to ﬁne sands. The proportion of each of these clast sizes deﬁnes the key differences between surface facies. Most of the sediment stored in the reach is potentially mobile. However, bankfull events or larger are required to mobilise some surface facies, as materials are stored in features that are elevated or distant from the low-ﬂow channel and the protective inﬂuence of vegetation is prominent. A conceptual model showing how the formation of sediment mixes for each surface facies relates to the distribution of geomorphic units, their elevation above the lowﬂow channel and different sized ﬂow events (or different stages of the same event) is presented in Fig. 16.7. The model has four sections; the bench (inundated in 5–10 year event), the main bar (inundated in 2–5 year event), the lower bar (inundated in 1–2 year event) and the rifﬂe (permanently inundated). Benches are the highest elevation units in the macrochannel and are most laterally distant from the low-ﬂow channel. They are very infrequently inundated and are likely to have been subjected to very little reworking since their initial formation. Benches are mainly composed of sands and granules, are very poorly sorted and are densely vegetated. They are generally composed of surface facies 1a, 1b and 2b. Surface facies 2b ranges in calibre from sand through to cobbles. For cobbles to be deposited at this elevation requires a bankfull event. These surfaces appear to have been deposited rapidly after periods of extensive bedload transport. There is no sorting of grains, similar to the sediment mixes observed in ephemeral streams (Laronne et al., 1994). Smaller events that inundate these surfaces may only be competent to deposit ﬁner sediments on the rising limb (surface facies 1b). In the waning stages of these events the ﬁne material carried in suspension settles in these low-velocity zones resulting in elevated, ﬂat and ﬁne-grained surfaces. This is the most common formation process associated with benches (Brierley and Fryirs, 2005) and explains why the majority of benches, and indeed most surfaces within the macrochannel are composed of surface facies 1a. Given its ﬁne-grained nature, this surface facies could be easily mobilised. However, height above and distance from the low ﬂow, as well as the presence

432

Very Infrequent Inundation 5-10 yr event Bench Zones 3, 6 & 9

1

Zone

3

Process

2

Facies

1

Floodplain

3&9

3&9

Dumped (larger event) 2b

Typical Unit s Bench

10

Dumped (smaller event) 1b Bench

6 Fine deposition (waning stages) 1a Bench

Bench Main Bar

6

Infrequent Inundation

Lower Bar

2-5 yr event Main Bar Area Zones 2, 5 & 8

5 4

10

Low Flow Channel

9

Zone

2, 5 & 8

Process

Dumped 2b

Facies

Typical Unit s Platform, Ridge

2, 5 & 8 Censored 3a Ramp, Platform

5&8 Fine deposition 1a Overflow Channel, Scour

2, 5 & 8 Infill/Embedding 2a Platform, Ridge Overflow Channel, Scour

8 7 Frequent Inundation

10

1 yr event

7

Lower Bar Area Zones 1, 4 & 7 Zone Process

Permanent Inundation

Facies

1&7 Censored 3a

Typical Unit s Upstream Sheet, Unit Bar, Inset Bar

4 Reworking 3c Apex Sheet

1&7

1& 7

Sheets of winnowed finer gravel 3b Inset Bar Upstream Sheet

1b Upstream Sheet, Unit Bar, Inset Bar

Zone Process Facies Typical Unit s

10

10

Censored

Reworked

3a Low flow

3c Riffle

Figure 16.7. Conceptual model relating the surface facies to elevation, inundation frequency and geomorphic unit location, with proposed formation processes shown.

J. Hoyle et al.

Riffle/Low Flow Channel Zone 10

Sediment organisation along the upper Hunter River, Australia

433

of often-dense vegetation cover, signiﬁcantly reduces its sensitivity to reworking. Establishment of dense vegetation between events protects these surfaces and the increased roughness promotes further deposition (Thornes, 1990; Tabacchi et al., 2000). The benches in the study reach appear to be undergoing sustained vertical accretion. The main bar areas are 2–3 m above the low-ﬂow channel and are only infrequently inundated (2–5 year events). They tend to have a mix of sizes, with calibre ranging from cobbles to sand (surface facies 2b). However, at these elevations, the duration and frequency of inundation and reworking is greater than on the bench, resulting in a range of geomorphic units. Areas where there is prolonged high-velocity ﬂow, such as ramps and platforms, are likely to result in surfaces devoid of ﬁne-grained materials (surface facies 3a). In the waning stages of events, ﬁner sediments may be deposited (surface facies 1a), particularly in areas where water is left to pool, such as scour holes and overﬂow channels (McGowen and Garner, 1970). In some areas these ﬁne-grained materials settle into the interstices embedding the larger clasts (surface facies 2a). Lower bar areas are more frequently inundated (1–2 year events) and therefore more frequently reworked. Their surfaces comprise coarse patches reworked sufﬁciently to remove ﬁner particles from the surface and become slightly imbricated or packed (surface facies 3c). Hydraulic modelling suggests that surface facies 3c is generally inundated in a 1–2 year event but that mobilisation of the D84 may require bankfull ﬂow. These periods of inundation may have provided the opportunity to mobilise the ﬁner grains resulting in the surface segregation and formation of a stable armour. These areas are subsequently immobile during smaller events. Surface facies 3c is the only facies to show this degree of organisation. The winnowed material forms bedload sheets of varying sizes depending on the calibre of material the ﬂow is competent to carry. Sheets range from cobble (surface facies 3a) to pebble (surface facies 3b) to granule and sand (surface facies 1b). These sheets are essentially the active mobile sediment and exhibit organisation most like that described in the gravel-bed rivers literature. Fine-grained material is also left in the lee of obstacles where secondary circulation cells form (Sambrook Smith and Ferguson, 1996). At this elevation the highest velocity ﬂow is on the apex of the bend. This explains why apex sheets are often coarser than upstream sheets or the ﬁne-grained inset bars. Fine-grained materials found at these low elevations are generally loose (due to recent deposition and lack of baking). The riffle and low-flow areas also contain sheets of censored pebbles and cobbles but the constant ﬂow inhibits deposition of ﬁne-grained materials on the surface. However, hyporheic ﬂows may deposit interstitial ﬁne-grained materials subsurface (Brunke and Gosner, 1997; Hancock and Boulton, 2005).

7.3.

How this model differs from freely adjusting rivers

Patterns of bed-material organisation along contemporary gravel bars of the Upper Hunter River differ from those described in the literature for frequently reworked bars in freely adjusting gravel-bed rivers. Of particular note is the remarkable lack of lateral or downstream ﬁning (within each bar and along the entire 8 km reach).

J. Hoyle et al.

434

Also, particle clusters and other surface structures are rare. Imbrication is very localised, and only slight in those patches in which it is observed. The channel in the study reach is not freely adjusting. Unlike simple point or lateral bars described in the gravel-bed literature, bar locations are forced by constrictions in the macrochannel, and bars are unable to migrate either laterally or longitudinally. The complex bar forms are also a result of the nature of channel adjustment following European settlement and the increased geomorphic effectiveness of major events, such as the 1955 ﬂood. These complex bars are characterised by a wide range of sediment sizes with a wide range of packing, censoring, baking and vegetation cover. While processes occurring in this river are not necessarily different to those occurring in gravel-bed rivers elsewhere, the complex array of surfaces, variable (in)frequency of reworking, and profound adjustment to sediment availability in the period since European settlement present a notable contrast to the boundary conditions that fashion more orderly distributions of bed material that have been characterised for ‘typical’ gravel-bed rivers. 7.3.1.

Controls on sediment availability

The Upper Hunter River ﬂows atop gravel substrates deposited under past higher energy conditions. The channel is conﬁned between cohesive, vertically accreted sediments, as well as bedrock and remnant terrace materials. These conditions limit the degree to which the river is able to freely adjust. They result in a low degree of withinreach reworking of bedload. The study reach also has a limited long-term supply of bedload from the upper catchment. Hence, relative to freely adjusting gravel-bed rivers in more tectonically active and/or glaciated regions, the Upper Hunter has very low-sediment availability. 7.3.2.

Discharge relationships

Signiﬁcant variability in ﬂow regime results in long periods of relative inactivity punctuated by occasional high magnitude-low frequency events that mobilise and rework bed materials of the Upper Hunter River. These high-magnitude bed mobilising events tend to be of short duration, usually being no more than a few days. Prior to European settlement, the energy of these high-magnitude events would have been readily dissipated over the ﬂoodplain, due to the small channel capacity, low-bed slope and high-instream roughness. The construction of Glenbawn Dam has resulted in increased low ﬂows and reduced peak ﬂows. Some elevated surfaces within the macrochannel may not have been reworked since formation, and may have experienced multiple depositional episodes. The infrequency of events enables vegetation to establish, signiﬁcantly reducing sediment mobility (Thornes, 1990). These conditions, together with the reduction in replenishment of bedload materials both from upstream and within-reach, are the key controls on the observed distribution of facies types. 7.3.3.

Impact of human disturbance since European settlement

The availability of sediments of different sizes has changed signiﬁcantly following impacts of human disturbance since European settlement. The simultaneous

Sediment organisation along the upper Hunter River, Australia

435

reduction of instream roughness and vegetation-induced bank cohesion (Brooks and Brierley, 2000) resulted in up to a threefold channel expansion in this reach. As a consequence, large volumes of ﬁne sediment were liberated from ﬂoodplain deposits along with appreciable, but lesser, volumes of gravels. Floods of between 10 and 20 year ARI are now contained within the channel zone, concentrating much greater proportions of the ﬂood energy within the channel than would have been the case pre-expansion. As ﬂows now rarely inundate the ﬂoodplain, a signiﬁcant proportion of the increased suspended sediment load that resulted from the channel expansion phase is now redeposited on the various surfaces within the macrochannel. Furthermore, there is a positive feedback between vegetation establishment on these surfaces and increasing washload deposition (Abt et al., 1994). Incision and expansion into the coarse gravel lag has also made coarse basal clasts (cobbles and boulders) available for potential mobilisation. However, apart from a period of intense ﬂood activity in the mid-20th century (Erskine and Warner, 1988), there have subsequently been few ﬂows competent to mobilise and organise these materials. The enlarged post-European channel capacity and the resultant concentration of ﬂood energy within the channel zone provide greater potential for vertical segregation of depositional units and greater potential for a diversity of geomorphic units. This, coupled with a highly variable ﬂow regime, provides increasing opportunity for vegetation colonisation on the diverse set of surfaces. This, in turn, induces a positive feedback mechanism whereby deposition of ﬁne-grained materials is enhanced, increasing stability (Brooks and Brierley, 2002). This partly explains the dominance of ﬁne-grained facies on many of the higher in-channel surfaces and the infrequency of their reworking. Within-reach variability in channel geometry and the assemblage of geomorphic units determine local-scale variability in sediment availability and the ease with which materials are prone to reworking. The benches are often elevated more than 4 m above the low-ﬂow surface. They are formed by infrequent large events and are poorly sorted in a fashion similar to that observed in ephemeral streams (Laronne et al., 1994). The main bars are generally made up of wide platforms elevated up to 3 m above the low-ﬂow surface and dissected by ridges, chutes and scour holes. These surfaces are more similar to those observed by Blum and Salvatore Valastro (1989) in Texan rivers. The inset lower bars sit just above the low-ﬂow surface and are much more mobile. These are the only surfaces in the reach that exhibit any patterns of longitudinal sediment sorting due to the presence of bedload sheets which are frequently reworked (Iseya and Ikeda, 1987; Whiting et al., 1988).

7.4. 7.4.1.

Implications Flux

Over engineering or management timescales, further incision or expansion along the reach is unlikely due to bedrock controls that act as a local base level. The main source of bedload for future ﬂux relies on the reworking of the bars within the reach as well as the background supply from upstream tributaries not impacted by dam impoundment.

J. Hoyle et al.

436

However, in its present condition, the reach is accumulating and storing materials within expanding benches in the zones widened during the post-colonial adjustment phase. This evolutionary inheritance fundamentally controls the bed-material organisation within this passive margin river, particularly when coupled with the characteristically variable ﬂow regimes within this region (Erskine and Warner, 1988; Finlayson and McMahon, 1988). Relatively infrequent, low-duration ﬂows are only potentially capable of transporting the gravel bedload on timescales in the order of days per decade. This is in marked contrast to the snow melt driven or glacially fed gravel-bed rivers of Europe, Central Asia, North America and New Zealand, in which ﬂows capable of transporting bedload often have durations of weeks or months annually. Thus, both sediment supply and the duration of bed-mobilising ﬂows are important controls on sediment organisation (Costa and O’Connor, 1995). At the other extreme, the gravel-bed desert streams described by Laronne et al. (1994) probably have even more irregular and shorter duration bed-mobilising ﬂows than those described here, and they appear to show an equivalent lack of sediment organisation in the mobile bars. Although there have been a number of events capable of mobilising and reordering the sediments made available since channel expansion on the Upper Hunter River, they have not resulted in patterns of sediment organisation that approximate those shown for freely adjusting meandering rivers. Given the elevation of the main bars and benches, they can only be reworked during high-magnitude events. Hence, storage times vary markedly for different sediment mixes. It appears that there are two modes of bedload transport occurring in the Upper Hunter. The ﬁrst involves mobilisation and reworking of material in the lowﬂow channel and on lower bar areas at ﬂow stages less than the 1–2 year event. At this scale the river is essentially freely adjusting as the low-ﬂow channel can shift to some degree within the macrochannel. As these bedload materials are frequently reworked, they are relatively well sorted and exhibit organisation similar to that described for freely adjusting gravel-bed rivers. The second mode involves the mobilisation of the main bar and bench areas. This requires much larger ﬂows, due to the elevation and calibre of material on some of these surfaces, and it is here that limitations of existing notions of bed mobility and bedload transport become apparent. 7.4.2.

Modelling

Quantiﬁcation of sediment ﬂux and surface sediment mobility in the low-ﬂow channel and lower bar areas relative to the main bar and bench areas must be considered as fundamentally different problems. Not only does each surface have a unique ﬂow recurrence interval, the different sediment mixes and vegetation communities on individual surfaces ensures that there are highly complex mobilisation relationships for each surface. Each unit has its own entrainment function and discharge/roughness relationship. Modelling applications must allow for this. Hence, sediment transport models based on 1D ﬂow hydraulics and reach-averaged roughness and sediment size characteristics will be very poor predictors of actual sediment transport for these kinds of rivers. No sensible D50 value can be used for all events. It may be possible to model small events where sediment mixes are more predictable and topography and patchiness less convoluted. However, as soon as multiple surfaces become inundated,

Sediment organisation along the upper Hunter River, Australia

437

patches of varying sediment size, sorting, packing and vegetation cover come into play, and the resultant ﬂow paths and erosion and deposition processes become highly complicated. In these circumstances, reach-scale spatially distributed 2D models are better equipped to adequately parameterise the hydraulics and sediment transport mechanics provided there is adequate input data available. 8.

Conclusion

The use of multivariate statistics, utilising a range of relevant bed parameters, provides a useful tool for characterising the nature of bedload materials that make up the ‘visible’ surface gravel fraction in the study reach. Facies development identiﬁed by these quantitative methods reveals relations between bed-material character and elevation above the thalweg that reﬂect geomorphic history and river responses to human disturbance. Gravel organisation along bars of the Upper Hunter River is notably different to that shown for gravel-bed rivers in the literature, with a lack of down-bar or down-reach ﬁning. Historical river changes have resulted in pronounced within-reach variability in the nature and extent of gravel surfaces. Over time, the connectivity between the channel and ﬂoodplain has been altered because of channel expansion, affecting the concentration of ﬂow energy within the channel. In addition, the nature and rate of bedload input into the reach and the cohesive, ﬁne-grained (ﬁne sand and silt) composition of the banks have inﬂuenced patterns of gravel organisation. Complex feedbacks related to vegetation cover on different geomorphic surfaces have signiﬁcant implications for future bedload mobility and sediment ﬂux. At present, existing bedload transport theory does not adequately address spatial variability in the availability of materials of differing character that may be mobilised at differing ﬂow stages. Considerable care is needed when using simple 1D models in settings where bedload mobility is an episodic, infrequent process that is confounded by numerous geomorphic controls, the consequences of which vary at differing spatial and temporal scales.

Acknowledgements We thank the following people who have helped pull this paper together by offering support, ﬁeld help, insight and expertise: Sarah Mika, Dan Keating, Mark Sanders, Tim Cohen, Hugh Jones, John Spencer and Simon McKee. Stephen Rice, Paolo Billi and Herve´ Pie´gay provided very valuable and constructive comments during the review process. Particular thanks go to Herve´ Pie´gay, Dave Nipperess and Hugh Jones for their advice regarding the statistical analysis used. This work is funded by an Australian Research Council Linkage Grant. References Abt, S.R., Clary, W.P., Thornton, C.I., 1994. Sediment deposition and entrapment in vegetated streambeds. J. Irrig. Drain. Eng. 120, 1098–1111.

438

J. Hoyle et al.

Ashmore, P.E., Parker, G., 1983. Conﬂuence scour in coarse braided streams. Water Resour. Res. 19, 392–402. Ashworth, P.J., Bennett, S.J., Best, J.L., McLelland, S.J. (Eds), 1996. Coherent Flow Structures in Open Channels. Wiley, Chichester. Ashworth, P.J., Ferguson, R.I., 1986. Interrelationships of channel processes, changes and sediments in a proglacial braided river. Geogr. Ann. 68A, 361–371. Ashworth, P.J., Ferguson, R.I., 1989. Size-selective entrainment of bed load in gravel bed streams. Water Resour. Res. 25, 627–634. Ashworth, P.J., Ferguson, R.I., Ashmore, P.E., et al., 1992. Measurements in a braided river chute and lobe. 2. Sorting of bedload during entrainment, transport and deposition. Water Resour. Res. 28, 1887–1896. Bartley, R., Rutherfurd, I., 1999. The recovery of geomorphic complexity in disturbed streams: using migrating sand slugs as a model. In: Rutherfurd, I. and Bartley, R. (Eds), Second Australian Stream Management Conference, Adelaide, pp. 39–44. Benzecri, J.P., 1992. Correspondence Analysis Handbook. Mercel Dekker Inc., New York. Best, J.L., Bristow, C.S. (Eds), 1993. Braided Rivers. Geological Society, London. Bluck, B.J., 1971. Sedimentation in the meandering River Endrick. Scott. J. Geol. 7, 93–138. Bluck, B.J., 1987. Bed forms and clast size changes in gravel-bed rivers. In: Richards, K. (Ed.), River Channel: Environment and Process. Blackwell, Oxford, pp. 159–178. Blum, M.D., Salvatore Valastro, J., 1989. Response of the Pedernales River of central Texas to late Holocene climatic change. Ann. Assoc. Am. Geogr. 79, 435–456. Brayshaw, A.C., Frostick, L.E., Reid, I., 1983. The hydrodynamics of particle clusters and sediment entrainment in coarse alluvial channels. Sedimentology 30, 137–143. Brierley, G., 1989. River planform facies models: the sedimentology of braided, wandering and meandering reaches of the Squamish River, British Columbia. Sediment. Geol. 61, 17–35. Brierley, G., Fryirs, K., 2005. Geomorphology and River Management: Applications of the River Styles Framework. Blackwell Publishing, Oxford, UK. Brierley, G., Miller, C., Brooks, A., et al., 2005. Making integrative, cross-disciplinary research happen: initial lessons from the Upper Hunter River Rehabilitation Initiative. In: Rutherfurd, I., Wiszniewski, I., Askey-Doran, M. and Glazik, R. (Eds), Proceedings of the Australian Stream Management Conference, Launceston, Tasmania, pp. 125–133. Brooks, A., Brierley, G., 1997. Geomorphic responses of lower Bega River to catchment disturbance, 1851–1926. Geomorphology 18, 291–304. Brooks, A., Brierley, G., 2000. The role of European disturbance in the metamorphosis of lower Bega River. In: Brizga, S.O. and Finlayson, B.L. (Eds), River Management: The Australasian Experience. Wiley, London, pp. 221–246. Brooks, A., Brierley, G., 2002. Mediated equilibrium: the inﬂuence of riparian vegetation and wood on the long term character and behaviour of a near pristine river. Earth Surf. Process. Landf. 27, 343–367. Brooks, A., Brierley, G., 2004. Framing realistic river rehabilitation targets in light of altered sediment supply and transport relationships: lessons from East Gippsland, Australia. Geomorphology 58, 107–123. Brooks, A., Brierley, G., Millar, R.G., 2003. The long-term control of vegetation and woody debris on channel and ﬂood-plain evolution: insights from a paired catchment study in southeastern Australia. Geomorphology 51, 7–29. Brunke, M., Gosner, T., 1997. The ecological signiﬁcance of exchange processes between rivers and groundwater. Freshw. Biol. 37, 1–33. Bufﬁngton, J.M., Montgomery, D.R., 1997. A systematic analysis of eight decades of incipient motion studies, with special reference to gravel-bedded rivers. Water Resour. Res. 33, 1993–2030. Burge, L.M., LaPointe, M.F., 2005. Understanding the temporal dynamics of the wandering Renous River, New Brunswick, Canada. Earth Surf. Process. Landf. 30, 1227–1250. Carling, P.A., Reader, N.A., 1982. Structure, composition and bulk properties of upland stream gravels. Earth Surf. Process. Landf. 7, 349–365. Church, M., Hassan, M.A., Wolcott, J.F., 1998. Stabilizing self-organized structures in gravel-bed stream channels: ﬁeld and experimental observations. Water Resour. Res. 34, 3169–3179.

Sediment organisation along the upper Hunter River, Australia

439

Clarke, K., Warwick, R., 2001. Change in Marine Communities: An Approach to Statistical Analysis and Interpretation. 2nd edn. PRIMER-E Ltd, Plymouth. Clement, P., Pie´gay, H., 2003. Statistics and ﬂuvial geomorphology. In: Kondolf, G.M. and Pie´gay, H. (Eds), Tools in Fluvial Geomorphology. Wiley, Chichester, pp. 597–630. Cohen, T., 2003. Late Holocene ﬂoodplain processes and post-European channel dynamics in a partly conﬁned valley of New South Wales, Australia. Unpublished Ph.D., University of Wollongong, Wollongong. Cohen, T., Brooks, A., Samosorn, B., 2004. The potential of airborne laser scanning (ALS) data to produce detailed surveys of rivers: an example from the Upper Hunter River, New South Wales, Australia. In: Rutherfurd, I., Wiszniewski, I., Askey-Doran, M., and Glazik, R. (Eds), Australian Stream Conference, Launceston, Tasmania, pp. 154–158. Costa, J.E., O’Connor, J.E., 1995. Geomorphically effective ﬂoods. In: Costa, J.E., Miller, A.W., Potter, K.W., and Wilcock, P.R. (Eds), Natural and Anthropogenic Inﬂuences in Fluvial Geomorphology, American Geophysical Union, pp. 45–56, Geophysical Monograph. Dawson, M., 1988. Sediment size variations in a braided reach of the Sunwapta River, Alberta, Canada. Earth Surf. Process. Landf. 11, 643–652. Desloges, J.R., Church, M., 1989. Wandering gravel-bed rivers. Can. Geogr. 33, 360–364. Dietrich, W.E., Smith, J.D., 1983. Inﬂuence of the point bar on ﬂow through curved channels. Water Resour. Res. 19, 1173–1192. Dietrich, W.E., Whiting, P.J., 1989. Boundary shear stress and sediment transport in river meanders of sand and gravel. In: Ikeda, S. and Parker, G. (Eds), River Meandering, American Geophysical Union Water Resources Monograph, Washington, DC, pp. 1–50. Digby, P.G.N., Kempton, R.A., 1987. Multivariate Analysis of Ecological Communities. Chapman and Hall, London. Erskine, W.D., Livingstone, E.L., 1999. In-channel benches: the role of ﬂoods in their formation and destruction on bedrock-conﬁned rivers. In: Miller, A.J. and Gupta, A. (Eds), Varieties of Fluvial Form. Wiley, New York, pp. 445–475. Erskine, W.D., Warner, R.F., 1988. Geomorphic effects of alternating ﬂood- and drought-dominated regimes on NSW coastal rivers. In: Warner, R.F. (Ed.), Fluvial Geomorphology of Australia. Academic Press, Sydney, pp. 223–244. Ferguson, R.I., Ashmore, P.E., Ashworth, P.J., et al., 1992. Measurements in a braided river chute and lobe, 1, ﬂow pattern, sediment transport, and channel change. Water Resour. Res. 28, 1877–1886. Finlayson, B.L., McMahon, T.A., 1988. Australia vs. the world: a comparative analysis of streamﬂow characteristics. In: Warner, R.F. (Ed.), Fluvial Geomorphology of Australia. Academic Press, Sydney, pp. 17–40. Fryirs, K., Brierley, G., 2001. Variability in sediment delivery and storage along river courses in Bega catchment, NSW, Australia. Geomorphology 38, 237–265. Fryirs, K., Brierley, G., Preston, N., Spencer, J., 2007. Catchment-scale (dis)connectivity in sediment ﬂux in the upper Hunter catchment, New South Wales, Australia. Geomorphology 84, 297–316. Gilvear, D.J., Cecil, J., Parsons, H., 2000. Channel change and vegetation diversity on a low-angle alluvial fan, River Feshie, Scotland. Aquat. Conserv. Mar. Freshw. Ecosyst. 10, 53–71. Greenacre, M.J., 1984. Theory and Applications of Correspondence Analysis. Academic Press Limited, London. Gurnell, A.M., Midgley, P., 1994. Aquatic weed growth and ﬂow resistance: inﬂuence on the relationship between discharge and stage over a 25 year river gauging station record. Hydrol. Process. 8, 63–73. Gustavson, T.C., 1978. Bed forms and stratiﬁcation types of modern gravel meander lobes, Nueces River, Texas. Sedimentology 25, 401–426. Hancock, P.J., Boulton, A.J., 2005. The effects of an environmental ﬂow release on water quality in the hyporheic zone of the Hunter River, Australia. Hydrobiologia 552, 75–85. Hassan, M.A., Reid, I., 1990. The inﬂuence of microform bed roughness elements on ﬂow and sediment transport in gravel bed rivers. Earth Surf. Process. Landf. 15, 739–750. HR Wallingford, 1990. Sediment transport, The Ackers and White Theory Revised, Report SR237, HR Wallingford, England.

440

J. Hoyle et al.

Ikeda, S., Parker, G. (Eds), 1989. River Meandering, American Geophysical Union Water Resources Monograph, American Geophysical Union, Washington, DC. Iseya, F., Ikeda, H., 1987. Pulsations in bedload transport rates induced by a longitudinal sediment sorting: a ﬂume study using sand and gravel mixtures. Geogr. Ann. 69A, 15–27. Jackson, R.G., 1976. Depositional model of point bars in the lower Wabash River. J. Sediment. Petrol. 46, 579–594. Karcz, I., 1972. Sedimentary structures formed by ﬂash ﬂoods in southern Israel. Sediment. Geol. 7, 161–182. Knighton, A.D., 1980. Longitudinal changes in size and sorting of stream-bed material in four English rivers. Geol. Soc. Am. Bull. 91, 55–62. Komar, P.D., Li, Z., 1988. Application of grain-pivoting and sliding analyses to selective entrainment of gravel and to ﬂow-competence evaluations. Sedimentology 35, 681–695. Laronne, J.B., Carson, M.A., 1976. Inter-relationships between bed morphology and bed material transport for a small gravel-bed channel. Sedimentology 23, 67–85. Laronne, J.B., Reid, I., Yitshak, Y., Frostick, L.E., 1994. The non-layering of gravel streambeds under ephemeral ﬂood regimes. J. Hydrol. 159, 353–363. Leopold, L.B., 1970. An improved method for size distribution of stream-bed gravel. Water Resour. Res. 6. Lisle, T., Ikeda, E.H., Iseya, F., 1991. Formation of alternate stationary bars in a steep channel with mixed size sediment: a ﬂume experiment. Earth Surf. Process. Landf. 16, 463–469. Lisle, T.E., Nelson, J.M., Pitlick, J., et al., 2000. Variability of bed mobility in natural, gravel-bed channels and adjustments to sediment load at local and reach scales. Water Resour. Res. 36, 3743–3755. Manly, B.F.J., 1991. Randomisation and Monte Carlo Methods in Biology. Chapman and Hall, London. McGowen, J.H., Garner, L.E., 1970. Physiographic features and stratiﬁcation types of coarse-grained point bars: modern and ancient examples. Sedimentology 14, 77–111. Nanson, G.C., 1986. Episodes of vertical accretion and catastrophic stripping; a model of disequilibrium ﬂood-plain development. Geol. Soc. Am. Bull. 97, 1467–1475. Nanson, G.C., Croke, J.C., 1992. A genetic classiﬁcation of ﬂoodplains. Geomorphology 4, 459–486. Nanson, G.C., Erskine, W.D., 1988. Episodic changes of channels and ﬂoodplains on coastal rivers in New South Wales. In: Warner, R.F. (Ed.), Fluvial Geomorphology of Australia. Academic Press, Australia, pp. 201–222. Nanson, G.C., Short, S.A., Price, D.M., 1992. Wetting and drying of Australia over the past 300 ka. Geology 20, 791–794. Nanson, G.C., Young, R.W., 1981. Downstream reduction of rural channel size with contrasting urban effects in small coastal streams of southeastern Australia. J. Hydrol. 52, 239–255. Page, K.J., Carden, Y.R., 1998. Channel adjustment following the crossing of a threshold: Tarcutta Creek, southeastern Australia. Aust. Geogr. Stud. 36, 289–311. Paola, C., 1989. Topographic sorting. EOS 70, 323. Parfait, A., 1998. Geomorphic Assessor. Department of Land and Water Conservation, Parramatta. Parker, G., Andrews, E.D., 1985. Sorting of bedload sediments by ﬂow in meander bends. Water Resour. Res. 21, 1361–1373. Parker, G., Klingeman, P.C., 1982. On why gravel bed streams are paved. Water Resour. Res. 18, 1409–1423. Pizzuto, J.E., 1995. Downstream ﬁning in a network of gravel-bedded rivers. Water Resour. Res. 31, 753–759. Powell, D.M., 1998. Patterns and processes of sediment sorting in gravel-bed rivers. Prog. Phys. Geogr. 22, 1–32. Reid, I., Frostick, L.E., 1984. Particle interaction and its effect on the thresholds of initial and ﬁnal bedload motion in coarse alluvial channels. In: Koster, E.H. and Steel, R.J.S. (Eds), Sedimentology of Gravels and Conglomerates. Canadian Society of Petroleum Geologists Memoir, Calgary, pp. 61–68. Rice, S.P., 1994. Towards a model of changes in bed material texture at the drainage basin scale. In: Kirkby, M.J. (Ed.), Process Models and Theoretical Geomorphology. Wiley, Chichester, pp. 159–172. Richards, K., 1982. Rivers: Form and Process in Alluvial Channels. Methuen, London. Richards, K., Clifford, N., 1991. Fluvial geomorphology: structured beds in gravelly rivers. Prog. Phys. Geogr. 15, 407–422.

Sediment organisation along the upper Hunter River, Australia

441

Ryan, P.D., Harper, D.A.T., Whalley, J.S., 1995. PALSTAT. Chapman and Hall, London. Sambrook Smith, G.H., Ferguson, R.I., 1996. The gravel-sand transition: a ﬂume study of channel response to reduced slope. Geomorphology 16, 147–159. Smith, J.D., McLean, S.R., 1984. A model for ﬂow in meandering streams. Water Resour. Res. 2, 1301–1315. Strom, K., Papanicolaou, A.N., Evangelopoulos, N., Odeh, M., 2004. Microforms in gravel bed rivers: formation, disintegration, and effects on bedload transport. J. Hydraul. Eng. ASCE 130, 554–567. Tabacchi, E., Lambs, L., Guilloy, H., et al., 2000. Impacts of riparian vegetation on hydrological processes. Hydrol. Process. 14, 2959–2976. Thioulouse, J., Chesel, D., Doledec, S., Olivier, J.-M., 1997. ADE-4: a multivariate analysis and graphical display software. Stat. Comput. 7, 75–83. Thorne, S.D., Furbish, D.J., 1995. Inﬂuences of coarse bank roughness on ﬂow within a sharply curved river bend. Geomorphology 12, 241–257. Thornes, J.B.E. (Ed.), 1990. Vegetation and Erosion. Wiley, Chichester. U.S.A.C.E., 2004. HEC-RAS, version 3.2.1. Hydrologic Engineering Centre, Davis. Van Niekerk, A.W., Heritage, G.L., Broadhurst, L.W., Moon, B.P., 1999. Bedrock anastomosing channel systems: morphology and dynamics of the Sabie River, Mpumulanga Province, South Africa. In: Miller, A.J. and Gupta, A. (Eds), Varieties of Fluvial Form. Wiley, Chichester, pp. 33–51. Van Sickle, J., 1997. Using means similarity dendrograms to evaluate classiﬁcations. J. Agric. Biol. Environ. Stat. 2, 370–388. Vietz, G., Stewardson, M., Rutherfurd, I., 2004. Not all benches are created equal: proposing and ﬁeld testing an in-channel river bench classiﬁcation, with implications for environmental ﬂows. In: Rutherfurd, I., Wiszniewski, I., Askey-Doran, M., and Glazik, R. (Eds), 4th Australian Stream Management Conference, Launceston, Tasmania, pp. 629–635. Whiting, P.J., Dietrich, W.E., Leopold, L.B., et al., 1988. Bedload sheets in heterogeneous sediments. Geology 16, 105–108. Wiberg, P.L., Smith, J.D., 1987. Calculations of the critical shear stress for motion of uniform and heterogeneous sediment. Water Resour. Res. 23, 1471–1480. Wilcock, P.R., Southard, J.B., 1989. Bedload transport of mixed size sediment: fractional transport rates, bedforms and the development of a coarse surface layer. Water Resour. Res. 25, 1629–1641. Williams, P.F., Rust, B.R., 1969. The sedimentology of a braided river. J. Sediment. Res. 39, 649–679. Williamson, W.H., 1958. Groundwater Resources of the Upper Hunter Valley, New South Wales. Govt. Printer, Sydney. Wittenberg, L., 2002. Structural patterns in coarse gravel river beds: typology, survey and assessment of the roles of grain size and river regime. Geogr. Ann. 84A, 25–37. Wolcott, J.F., 1989. Flume studies of gravel bed surface response to ﬂowing water. Unpublished, University of British Columbia. Wolman, M.G., 1954. A method of sampling coarse river bed material. Trans. Am. Geophys. Union 35, 951–956.

Gravel-Bed Rivers VI: From Process Understanding to River Restoration H. Habersack, H. Pie´gay, M. Rinaldi, Editors r 2008 Elsevier B.V. All rights reserved.

443

17 The evolution of sediment waves inﬂuenced by varying transport capacity in heterogeneous rivers Thomas E. Lisle

Abstract Large sediment inputs can cause pronounced variations in sediment storage and ﬂux in a drainage network and lead to deformation of alluvial forms and impacts on river resources. Recent research has attempted to isolate this problem by focusing on the evolution of individual sediment waves created by large point inputs. New onedimensional models can predict streamwise variation of sediment transport and deposition for engineering problems such as the fate of sediment released from decommissioned dams. Model predictions and supporting evidence from experiments and ﬁeld studies indicate that dispersion has a strong inﬂuence on wave evolution in gravel-bed channels. However, signiﬁcant translation can occur where the input material is substantially ﬁner than the ambient bed material, and streamwise variations in channel morphology common to many natural channels can imprint reachscale variations in storage otherwise absent in uniform channels. Three-dimensional patterns of erosion and deposition associated with sediment waves pose important unsolved problems for the dynamics of alluvial forms and the effects on aquatic and riparian ecosystems. Transport capacity mediates transfer and storage of sediment in channels responding to sediment disturbances. Over time scales of cycles of aggradation and degradation that are associated with sediment waves, transport capacity can be expected to vary with changes in channel morphology and bed texture. Degrading sediment reservoirs commonly exhibit a positive relation between sediment transfer rate and storage. However, results from a recent ﬂume experiment indicate that transport– storage relations during cycles of aggradation and degradation may not be constant, but vary due to lags in input and downstream transfer of sediment in a reservoir and changes in mobility between aggrading and degrading states. Filling of storage may contribute indirectly to increased transfer rates, but over short spatiotemporal scales, transfer rates appear to be more directly inﬂuenced by local sediment supply within the immediate channel. E-mail address: [email protected] ISSN: 0928-2025

DOI: 10.1016/S0928-2025(07)11136-6

T.E. Lisle

444 1.

Introduction

The downstream transfer of bed material is critical to understanding sediment routing and the evolution of alluvial landforms in gravel-bed river systems. Erosion and deposition of bed material, as well as bank accretion by suspended sediment, mold the surface that transfers all sizes of sediment through the system and from channel to ﬂoodplain. Interactions between ﬂow, sediment transport, and morphology mediate the signal of environmental change recorded in landforms and stratigraphy. Bed-material routing in gravel-bed channels also has important practical applications in predicting, for example, the redistribution of augmented gravel supplies below dams, the fate of sediment released from decommissioned dams, and the cumulative effects of watershed disturbances that are transmitted by increased sediment loads. The purpose of this paper is to review recent research of bed-material routing, which has received accelerated interest since the review of Nicholas et al. (1995), and to identify some critical outstanding problems. I ﬁrst discuss the evolution of bed-material waves, which enables one to focus on individual perturbations in the downstream transfer of sediment and thereby simplify the problem of bed-material routing. I then ﬂip the problem and consider variations in the capacity of reaches of river to transfer sediment downstream. The general concept is that large sediment inputs create disturbances in bed material transfer (sediment waves) that impose spatiotemporal variations in sediment supply on channels, which eventually respond by maintaining continuity. The response by sediment storage and transfer is mediated by the dynamics of transport capacity, which is constrained by channel form and texture and results in the morphodynamics of diverse alluvial landforms. I focus on bed-material transport and storage in gravel-bed channels during ﬂows when sediment concentrations are too low to substantially affect ﬂow properties (i.e., disregarding debris ﬂows and hyperconcentrated ﬂows). I also do not discuss network routing of sediment (but see Benda and Dunne, 1997).

2.

Evolution of sediment waves

The evolution of sediment waves is the outcome of interactions between channel morphology, sediment transport, and ﬂow. All of these inﬂuences vary in space and time due to mutual adjustments and external factors, e.g., hydrologic events. Thus the problem is complex enough without regarding selective transport of mixed-size sediment. My approach is to ﬁrst discuss the simplest case of one-dimensional waves in uniform channels and then add complications (and fewer solutions) in multidimensional cases. 2.1.

Wave definitions and scales

Sediment waves (pulses or slugs) are transient zones of sediment accumulation in channels that are created by sediment inputs (Lisle et al., 2001) and are the cause of cycles of aggradation and degradation (Nicholas et al., 1995). A sediment wave is not

The evolution of sediment waves in heterogeneous rivers

445

necessarily composed exclusively of the population of sediment particles of the original input, but more generally is a local disturbance in the storage and downstream transfer of sediment that results from a sediment input. This accommodates the common case where particles transported by the channel from upstream are incorporated in the wave. Particles downstream of the initial input can also be incorporated in the spreading wave, and wave particles can be put into long-term storage. Longitudinal variations in channel topography can modify and accentuate sediment waves during their spread or migration, but they cannot create them, e.g., a sedimentation zone (Church, 1983) is not in itself a sediment wave. Sediment pulses are zones of high transport rate (Reid et al., 1985; Iseya and Ikeda, 1987), which for this discussion are associated with sediment waves and regarded at a commensurate scale. Terms for these features are problematic. The ﬁrst to describe them was Gilbert (1917) who used the term ‘sediment wave’ by analogy to a ﬂood wave. Nicholas et al. (1995) chose ‘sediment slug’ based on the uncertainty of transport processes replicating ‘true wave motion’. The complexity of processes involved in evolution of these features, as described below, bears this out. However, ‘slug’ seems to imply that the feature is exclusively the body of introduced sediment and does not include ambient sediment affected by the disturbance. Perhaps for both reasons, Cui and Parker (2005) chose ‘sediment pulse’, but this conﬂicts with earlier usages referring to variations in sediment transport rates. For lack of a clear alternative and in deference to Gilbert, I choose the term ‘sediment wave’ for an excess of bed material and ‘pulse’ for a zone of pronounced bed-material transport. Sediment waves are large-scale features. Nicholas et al. (1995) provide a classiﬁcation of sediment slugs (waves) based on a parallel classiﬁcation of bedforms (after Jackson, 1975; Church and Jones, 1982; Hoey, 1992) (Table 17.1), and list some examples. All but some macroslugs persist for time scales longer than a hydrologic event. Some superslugs have existed for over a century (Trimble, 1981; James, 1991) and others in earlier stages of development are likely to persist for a number of decades or longer (Madej and Ozaki, 1996; Trustrum et al., 1999; Bartley and Rutherford, 2005). Characteristic length scales range from 100 to 102 channel widths. Easily discernable waves are not common in nature (but see Pickup et al., 1983; Meade, 1985; Knighton, 1989; James, 1991; Madej and Ozaki, 1996; Miller and Benda, 2000; Sutherland et al., 2002; Kasai et al., 2004; Korup, 2004; Bartley and Rutherford, 2005), because they have to be large to be detectable (Lisle et al., 2001) and separated from other waves produced during the same triggering event. It is more common to have pervasive or punctuated but well-distributed sources of sediment that amalgamate in a river network (Gilbert, 1917; Knighton, 1989; Madej and Ozaki, 1996; Trustrum et al., 1999; Bartley and Rutherford, 2005; James, 2006).

2.2.

Translation and dispersion

Understanding the relative tendencies for translation or dispersion of sediment waves is important to explaining many aspects of river behavior and the sedimentary record associated with deposition and remobilization of bed material, particularly in rivers that receive large inputs of sediment. It also has practical value for predicting

446 Table 17.1. Parallel classiﬁcations of bedforms and sediment slugs (adapted from Nicholas et al., 1995) and publications since 1995 describing natural examples. Bedform class

Features

Equivalent slug class

Post-1995 examples

Dominant controls

Impact on ﬂuvial system

Macroforms

Bars

Macroslug

Fluvial processform interactions

Minor channel change

Megaforms

Bar assemblages

Megaslug

Local sediment supply and valley-ﬂoor conﬁguration

Major channel change

–

–

Superslug

Lane et al. (1996); Wathen and Hoey (1998) Turner (1995); Miller and Benda (2000); Wohl and Cenderelli (2000); Sutherland et al. (2002); Kasai et al. (2004) Madej and Ozaki (1996); Korup (2004)

Basin-scale sediment supply

Major valley-ﬂoor adjustment

T.E. Lisle

The evolution of sediment waves in heterogeneous rivers

447

downstream effects of sediment inputs on ﬂood risk, channel instability, and riverine habitats. Along channels inﬂuenced by sediment waves, dispersion tends to mute and spread sedimentation effects; translation tends to maintain the amplitude and concentration but limit the duration of sedimentation effects. Knowledge of wave behavior is commonly uncertain because of the lack of data on channel topography before the introduction of a wave. Some inferences can be drawn from the general behavior of gravel-bed channels. The signature of strong translation of sediment waves would be pronounced variations in bed elevation, transport rate, and channel morphology. Observations of rivers receiving large sediment inputs are telling for what they rarely report: common, unambiguous examples of waves of bed material migrating downstream. Deposits left by sediment waves tend to be thickest near the point of origin and thin downstream (Pickup and Higgins, 1979; James, 1991; Turner, 1995). Although these indicate a strong tendency for dispersion, a closer look is warranted to test for signiﬁcant translation. Wave translation in gravel-bed channels has been tested using theory (Wathen and Hoey, 1998; Cao and Carling, 2003, 2005; Cui et al., 2005), ﬂume experiments (Lisle et al., 1997; Cui et al., 2003a), and ﬁeld studies (Miller and Benda, 2000; Sutherland et al., 2002; Kasai et al., 2004). Theory and ﬂume experiments have deﬁned limitations on translational behavior in simple, uniform channels with ﬁxed banks (Lisle et al., 1997; Cui et al., 2003a,b), but application of these predictions to a variety of natural channels remains uncertain because of a lack of testing of comprehensive ﬁeld evidence against unambiguous criteria for translation. Criteria for a translating wave are the downstream migration of the wave apex and the trailing limb (Lisle et al., 2001). The leading edge advances in both dispersing and translating waves. Detecting translation can be difﬁcult because the amplitude of a sediment wave must decay to less than the amplitude of bar-pool sequences before translation is possible. During initial stages, the backwater effect of a wave not only prevents erosion of the trailing edge, but also causes the wave to extend upstream by deposition of incoming bedload. This effect disappears once upstream deposition and downstream erosion ﬂatten the apex to a fraction of bankfull depth and allow erosion of the trailing limb, thus enabling translation if other conditions for translation exist (Cui et al., 2003b). At this low amplitude, bar-scale variations in bed elevation make it difﬁcult to distinguish deviations from the pre-wave longitudinal proﬁle. Tracking variations in sediment character such as particle size, lithology, or degree of weathering can be misleading because a sediment wave includes more particles than those of the initial input. Therefore, bed elevation along the pathway of the wave must be known accurately in order to distinguish translating from non-translating waves. Such data are rare and the investigator must reconstruct pre-wave data from limited information and simple assumptions (e.g., a straight pre-wave proﬁle or depth to bedrock). Under these circumstances, interpretation of longitudinal proﬁles can be ambiguous. Lisle et al. (2001) argued from a simple geometric analysis that dispersing waves and translating waves produce similar lagged rises and falls in bed elevation downstream. The importance of accurate pre-wave proﬁles can be illustrated with another geometric analysis. Consider changes in hypothetical exponential proﬁles extending downstream from the apex of a dispersing wave in which mass is conserved and no material is added from upstream (Fig. 17.1). Without knowledge of pre-wave bed elevations or

448

T.E. Lisle

Figure 17.1. (A) Evolution of a sediment wave as a series of lengthening exponential proﬁles extending downstream of the apex and conserving mass. Perceived evolution of a sediment wave is based on two proﬁle datums: (B) the earliest post-input proﬁle; (C) the proﬁle of minimum elevations measured postinput. Elevation and distance scales are arbitrary.

The evolution of sediment waves in heterogeneous rivers

449

the basis to extrapolate accurately beyond the affected channel, an investigator might choose between two reference proﬁles to analyze wave evolution. One case uses the earliest wave proﬁle available [e.g., from the highest remnant terrace (Beschta, 1983)]; the other uses the lowest elevations surveyed during the evolution of the wave, assuming that as bed elevations are monitored longer, they will approximate pre-wave values more closely. Wave evolution is measured by comparing successive surveys with the reference proﬁle. In the ﬁrst case, there is an initial jump in the apex and trailing edge downstream, followed by translation of both trailing edge and apex of the exponential proﬁle. The second case produces a compound wave with a ﬁxed peak elevation at the origin and a downstream lobe that appears to have been shed from the upstream portion and progresses downstream. Thus, both examples show signs of translatory behavior where none exists. An alternative in detecting the limits of the wave from topographic surveys is to examine longitudinal variations of morphologic features of the channel that respond to the presence of increased bed material. These include channel width, pool depth, variability in thalweg elevation, median particle size, particle sorting, sediment thickness, and bedrock exposure (Wohl and Cenderelli, 2000; Kasai et al., 2004; Bartley and Rutherford, 2005). Though indirect, these methods can indicate the presence of excess bed material in the usual cases where pre-wave data on bed elevation are unavailable.

2.3.

Analysis: sediment waves in one-dimensional, uniform channels

Longitudinal changes in many factors inﬂuencing erosion and deposition, e.g., channel morphology, gradient, and tributary inputs, can affect the evolution of sediment waves in natural rivers. A theoretical entry to the problem is to model interactions of ﬂow, sediment transport, and channel morphology in one dimension under steady ﬂow (Lisle et al., 1997; Dodd, 1998; Cui and Parker, 2005). More comprehensive models add unsteady ﬂow, mixed particle-size distributions, and non-uniformity in channel width and gradient (Cui and Parker, 2005; Cui et al., 2006). To date, fully dimensional models are not adequately supported by theory and would present formidable computational requirements. In an attempt to provide a physical explanation of wave evolution, Lisle et al. (2001) use an analytical model based on one-dimensional equations for ﬂow continuity, ﬂow momentum, and sediment continuity to express factors inﬂuencing relative degrees of wave dispersion and translation. The model is simpliﬁed by incorporating a + form of the Meyer–Peter–Muller transport equation (Lisle et al., 1997; Dodd, 1998), but similarly based numerical models that have accurately predicted wave evolution in rivers (Cui et al., 2003b; Cui and Parker, 2005) incorporate other transport equations. The model of Cui and Parker (2005) considers the transport and conservation of heterogeneous sediment, particle abrasion, varied ﬂow, and non-uniform width. Another version (Cui et al., 2006) is used to predict the downstream redistribution of sediment released from decommissioned dams. Solutions derived from these models indicate that dispersion is prevalent in all cases and translation is signiﬁcant only in cases of low Froude number, as long as sediment concentrations are not extreme and

T.E. Lisle

450

input material is not ﬁner than ambient bed material. Gravel-bed rivers tend to have high Froude numbers at channel-forming discharge in comparison to sand-bed rivers. Elements of the theory presented by Lisle et al. (2001) are contested (Cao and Carling, 2003, 2005) and defended (Cui et al., 2005). Disregarding inﬂuences of Froude number, effects of wave topography on wave evolution can be conceptualized with a simple model. Let us assume that sediment transport tends to decrease and the bed to aggrade in zones of decreasing boundary shear stress (approximated by local gradient); conversely, transport tends to increase and the bed to degrade in zones of increasing shear stress. Consequently, deposition is favored where the bed shoals on the upstream limb, erosion is favored immediately beyond the wave crest, and deposition is again favored where the downstream limb tapers into the pre-existing channel. This combination would promote a spreading of the limbs of the wave and a ﬂattening of the crest, or overall dispersion of the wave. These analyses suggest that interactions between ﬂow, transport, and topography alone do not lead to signiﬁcant translation, except in channels with low Froude number, which does not typify most gravel-bed rivers. Moreover, additional inﬂuences that cause differential transport of particles, including selective transport and transverse variations in scour and deposition, promote dispersion.

2.4.

Streamwise variations in transport rate

Dispersion of the mass of a sediment wave does not preclude translation of other sedimentary conditions. Streamwise variations in transport rate could be much more evident and inﬂuential on river resources and ecosystems than a modest change in bed elevation. Downstream progression in peak transport rates would give the impression of wave translation even if the mass of a wave were to disperse without translation. In a laboratory channel, local bedload transport rates during wave evolution were as much as two orders of magnitude greater than those measured before the wave was introduced and the channel was at equilibrium; peak transport rates progressed downstream over an exclusively dispersing wave (Cui et al., 2003a). The downstream translation of peak transport rates is a necessary outcome of both dispersing and translating waves. Intuitively, and as expressed in the Exner equation, ð1 lÞ

@z @qs ¼0 þ @t @x

(17.1)

(where l is sediment porosity, z the bed elevation, t the time, qs the transport rate, and x the downstream distance), transport rates increase downstream in zones of erosion (e.g., beyond the wave crest) and decrease in zones of deposition (e.g., over the downstream prograding limb). The hinge point between erosion and deposition in a channel without lateral inputs marks the zone of maximum transport. The hinge point of a solely dispersing wave could migrate downstream slowly or remain nearly ﬁxed depending on developing wave proﬁles; the hinge point of a translating wave would show marked downstream migration. Hinge points can be observed from longitudinal surveys of a reach of the Navarro River in northern California that contains a dispersive sediment wave (Fig. 17.2). The wave was created

The evolution of sediment waves in heterogeneous rivers

451

Figure 17.2. Longitudinal proﬁles of a reach of the Navarro River in northern California, 1995–1997.

by a landslide that partially dammed the river in 1995 (Sutherland et al., 2002). In the next two years, the hinge point translated approximately 300 m as the downstream limb of the wave advanced and the apex degraded in place. Other examples of migrating hinge points on landslide-induced, dispersive waves are provided by Turner (1995) and Brummer and Montgomery (2006). 2.5.

Influence of particle size

Large sediment inputs are rarely of the same particle size as the bed material of a receiving channel, and are commonly ﬁner. Erosion of soil mantles in granitic terrain, for example, can contribute mostly sand to channels armored with gravel and boulders (Benda et al., 2004). Results from experiments and case studies of natural rivers show a strong inﬂuence of the relative particle size of wave material on wave behavior: coarser material generally slows wave evolution and enhances dispersion; ﬁner material accelerates wave evolution and promotes translation. Large input particles tend to form coarse armor layers, thereby decreasing transport rates and causing ﬁner particles to be selectively transported farther downstream than coarser particles (Cui and Parker, 2005; Brummer and Montgomery, 2006). In some cases a lag deposit is created near the point of entry (Turner, 1995; Brummer and Montgomery, 2006). Intensive transport of ﬁne material that is contributed to coarse-bedded channels can cause rapid evolution of sediment waves. For example, debris ﬂows that were generated by thunderstorms deposited large volumes of sand at the mouths of tributaries of the North Boise River, Idaho, which is generally well armored by cobbles and boulders (Benda et al., 2004). Within a single high-runoff season, these sandy waves spread to a reservoir 50 km downstream and dispersed to the level whereby armor particles were everywhere exposed (Gott, 1998). Wave translation requires erosion of the upstream limb of the wave, which can occur in a ﬁne-grained wave once it initially disperses and backwater effects become vanishingly small. This enables selective transport of ﬁner material over the ambient

T.E. Lisle

452

bed material and downstream migration of the upstream limb. One of the clearest examples of this phenomenon is provided by the East Fork River, Wyoming, where irrigation-return ﬂow in the Muddy River contributes a wave of sand to the East Fork, an armored gravel-bed river, each summer when ﬂow in the East Fork is low (Meade, 1985). During the subsequent nival ﬂood, each year’s wave disperses but maintains its coherence as it advances downstream. Fine-grained waves were modeled in Run 4B by Cui et al. (2003a), in which a wave of sand was laid onto an armored gravel bed that was formed by feeding a gravel–sand mixture at low rates over a long period. This and previous runs with gravel waves which did not translate were conducted with the same slope, discharge, and initial bed conditions. Within 2 h, the sandy wave in Run 4B dispersed and translated until the trailing edge of sand exited the ﬂume. In contrast, gravelly waves composed of the same mixture as the feed dispersed and disappeared in a period of 30 h. The evolution of all waves was predicted reasonably well with the model (Cui et al., 2003b). Table 17.2 includes studies where initial conditions and governing factors of evolving waves are well documented. By happenstance, cases where wave material is distinctly ﬁner than ambient bed material cover exactly the range of Froude number as those where particle sizes of the wave and the bed are equal or nearly so. Waves with ﬁner material exhibit translation as well as dispersion; waves with equivalent material exhibit dispersion only. In the absence of case studies with equivalent material that exhibit translation, the most deﬁnitive conclusion that can be made is that an upper limit of Froude number promoting translation independent of a grain-size effect is less than 0.4. A third case exhibiting translation and dispersion includes sandy waves moving over immobile beds (clay) in rivers with low Froude number.

2.6.

Influence of abrasion

Sediment inputs also commonly differ from ambient bed material with respect to abrasion rate. Landslide material, for example, can include fractured and weathered bedrock and regolith that wears rapidly (Jones and Humphrey, 1997; Sutherland et al., 2002), while ambient gravel particles traveling long distances tend to be selected for greater resistance (Werritty, 1992). Landslide material entering the Navarro River had a laboratory-measured abrasion rate of 0.5–1.0 km1, causing approximately one-half to abrade to the size of suspended sediment before being transported 1 km, and thereby lost from the bed-material wave (Sutherland et al., 2002). Simulations by Cui and Parker (2005) demonstrate that high abrasion rates of input material can cause the center of mass of a sediment wave to shift upstream as backwater effects cause progressive deposition of ambient bedload upstream, and input material abrades and disappears before it can advance far downstream. Conversely, hard particles in sediment inputs can increase the life span of a wave and allow it to accrete more ambient material. Their simulations show that the effect of abrasion on wave evolution becomes minimal when abrasion coefﬁcients (aw) are below 0.01 km1 (aw in Sternberg’s (1875) relation, M x =M 0 ¼ e awx , where Mx ¼ mass of bedload particles after travel distance x; M0 ¼ mass of bedload particles at the starting point). This value is not particularly low for many sedimentary rocks.

Governing conditions and behavior of well-documented sediment waves.

Source

River East Fork Navarro Fall R, CO NF Poudre R, CO Ringarooma Creighton Wannon Flume U Tsukuba SAFL-Run 2 SAFL-Run 4b

Froude

Sediment sizesa

Behaviorb

References

Wave

Bed

0.36 0.36 0.5 0.50 0.36 0.3 0.16

s gr-s s s s s s

gr gr-s gr gr gr clay clay

T/D D T/D T/D T/D T/D T/D

Meade (1985) Sutherland et al. (2002) Pitlick (1993) Wohl and Cenderelli (2000) Knighton (1989), Bartley and Rutherford (2005) Bartley and Rutherford (2005) Bartley and Rutherford (2005)

0.9 0.8 0.9

s-gr s-gr s

s-gr s-gr s-gr

D D T/D

Lisle et al. (1997) Cui et al. (2003a) Cui et al. (2003a)

The evolution of sediment waves in heterogeneous rivers

Table 17.2.

Note: Froude values are computed from mean hydraulic variables: at bankfull stage for rivers and for constant discharge for ﬂumes. a s, sand; gr, gravel. b T, translation; D, dispersion.

453

T.E. Lisle

454 2.7.

Topographic forcing

Topographic forcing by zones of high or low transport capacity along the path of a sediment wave can induce differences in deposition and erosion downstream (Beschta, 1983; Church, 1983; Nakamura et al., 1995; Cui and Wilcox, 2005), causing aggradation in sedimentation zones and manifesting a form of wave translation where it would not occur under uniform conditions. Such sedimentation zones, which are longer than bar-pool sequences, are characterized by contrasts in gradient and conﬁnement. Given topographic variations in natural channels and the low amplitude of sediment waves, departures from a regular wave proﬁle over segments of 101 to 102 channel widths in length (megaform scale) can be expected. For example, Cui and Wilcox (2005) predict that as much as 1.5 m of aggradation would occur in a lowgradient reach 9 km downstream of the Marmot Dam on the Sandy River, Oregon, after dam removal and subsequent release of 750,000 m3 of gravel and sand, while deposition in the intervening gorge would be negligible. Whether translation is topographically forced or inherent to sediment waves in channels of a certain type may be merely a theoretical issue for a practitioner, but evaluation of the effect of large-scale non-uniformity in channels is necessary for accurate predictions of sedimentation associated with large waves.

2.8.

Some three-dimensional effects

Three-dimensional effects on downstream transfer of sediment are evident in the dispersive nature of deposition and the convergent nature of erosion. For example, river bars are constructed by diverging, sediment-bearing ﬂow and pools are scoured by converging ﬂow (Church and Jones, 1982). Likewise, widening and braiding are commonly associated with aggradation; channel narrowing and re-establishment of single-thread channels are commonly associated with incision (e.g., Lyons and Beschta, 1983). Moreover, lateral erosion of deposits commonly lags incision (Meade, 1982; Nakamura et al., 1995). Even if one-dimensional models can provide approximate ﬁrst-order predictions of channel response to sediment waves, a deeper understanding of these processes requires a three-dimensional perspective. An expected outcome of the different patterns of erosion and deposition is that the same sediment particles are not transferred in each cycle of deposition and remobilization, and exchanges between erosion and deposition cannot be expected to be exactly compensatory in volume or particle size, at least in the short term. Moreover, three-dimensional considerations of wave behavior are vital to understanding ecological consequences, because habitat patches for species and communities are commonly conﬁgured laterally as well as longitudinally. Migration of a sediment pulse into a sedimentation zone can trigger complex interactions between sediment ﬂuxes and bar instability that propagate pulse behavior (Beschta, 1983). Increases in sediment load tend to build bars and redirect ﬂows into erosive boundaries (Dietrich and Smith, 1983; Ashworth, 1996). Avulsion of the channel across a medial bar during a large ﬂood in Allt Dubhaig, Scotland, altered ﬂow patterns downstream, mobilized stored sediment, induced bar deformation, and

The evolution of sediment waves in heterogeneous rivers

455

increased the ﬂux of bed material downstream (Wathen and Hoey, 1998). In this case, an endogenous pulse or auto-wave (macro-form scale) was initiated by channel erosion from a large ﬂood, rather than by an extraneous sediment input. Deposition and erosion in three channel sections were not balanced, thus net ﬂux between sections for the duration of the pulse was unequal. In fact, the net result of the pulse was a depletion of stored material in the reach. These observations suggest that, as bedmaterial transport characteristically occurs from bar to bar in gravel-bed channels, a large-scale disturbance from a sediment wave is likely to be manifested at the advancing front as increased bar instability in channels with large volumes of bed material stored in active bars. Because sediment transfer between bars can be partially discontinuous, net deposition of pulse sediment on a bar could abruptly halt a pulse, or net erosion could locally amplify it. An example of wave dissipation comes from observations at the leading edge of a landslide-induced sediment wave in the Navarro River (Sutherland et al., 2002). Advancement of the leading edge of this dispersive wave caused deposition of less than 0.5 m on a central bar and lateral erosion of the right bank, which was 4 m high and consisted of silt and ﬁne sand. Although there was a net input of 300 m3 of sediment downstream, this did not contribute to the bed-material load, and widening of the reach that contains the central bar increased the bed-material storage capacity. Subsequent surveys of bed elevation have not shown advancement of the leading edge of the wave downstream of the central bar. Discontinuities in sediment transfer may be most pronounced in channels where bed-material load is low relative to storage capacity and sediment exchange between reservoirs storing different particle sizes is common (Hooke, 2003). For example, whether bank erosion leads to propagation or deﬂation of a bed-material wave would depend on the fraction of bank material contributing to the bed-material load and the height of the eroded bank (Neill, 1987). Large woody debris in small forested channels can serve as ‘wood valves’ to store and release bed material. For example, breakup of wood jams in concert with sediment inputs from upstream have created pronounced sediment waves in Carnation Creek (Hassan et al., this volume). Another outstanding problem is the inﬂuence of unsteady ﬂow. Sediment waves commonly persist over many hydrologic events and are inﬂuenced by a wide range of ﬂows that differ in their capacity to mobilize particle sizes and areas of the bed. Inputs of sediment and ﬂow can be decoupled even for small sediment waves in a reach of channel (Lane et al., 1996). Many sediment waves originate as large sediment inputs caused by large ﬂoods, which initially distribute the sediment extensively over alluvial surfaces, and are later reworked by lower ﬂows (e.g., Beschta, 1983; Madej and Ozaki, 1996; Miller and Benda, 2000). Selective transport and mobilization of narrow zones of bed material by such ﬂows can be expected to promote dispersion.

3. Transport capacity and storage: spatial and temporal patterns of transmitting changes in sediment supply As bed-material waves disperse over long reaches of gravel-bed rivers, variations in transport capacity due to non-uniform conditions force patterns in erosion and

456

T.E. Lisle

deposition that mask those inherent to wave behavior and evident under uniform conditions. This motivates ﬂipping the problem of sediment routing from the previous approach of examining wave behavior to a complementary approach of examining interactions between sediment storage and transport capacity under varying sediment supply. To analyze a non-uniform system, we can disassemble a river into quasi-uniform sediment reservoirs – valley segments that include channels, ﬂoodplains, and alluvial terraces (Lisle and Church, 2002). Each reservoir is distinguished from its neighbors by its particular capacity to store or transfer sediment, depending on the rate of sediment input and the volume of sediment stored. A sedimentary system can be modeled as a series of reservoirs linked by the transfer of sediment (e.g., Benda and Dunne, 1997). This approach focuses on ﬂuvial processes governing the dynamics of alluvial forms. Transport capacity at a given point along a drainage network appears constant or dynamic depending on whether the time scale of observation is equal to that of the passage of a sediment pulse (or episode of aggradation and degradation). Transport capacity can be regarded as constant for a time scale corresponding to basin evolution (Z103 yr), since in the long term, the channel must convey sediment at the rate of denudation, taking into account accretion onto internal basins. At this scale, changes in transport and storage associated with a sediment wave can be disregarded. Transport capacity can also be regarded as constant at a much shorter time scale corresponding to a hydrologic event (r100 yr), which is often less than that of the passage of a sediment pulse. At this scale, sediment is transported according to the range of ﬂows and the physical conditions of the channel that govern sediment transport. The operative relation between sediment transport and ﬂow represents a snapshot during the evolution of sediment transport conditions as rates of input and storage change. Transport capacity at the time scale of a hydrologic event is not necessarily equal to that representing basin evolution because of variable conditions governing channel mobility, even under a constant hydrologic regime. Transport capacity is dynamic at an intermediate time scale (100–102 yr) corresponding to the evolution of a sediment wave and associated changes in sediment storage in parts of a basin (Lisle and Church, 2002). At this scale, conditions governing sediment transport under the hydrologic regime respond to variations in storage and sediment ﬂux as sediment waves extend, overlap, and attenuate through the network. Sediment waves are easiest to analyze individually (Nicholas et al., 1995; Lisle et al., 2001), but multiple waves from tributaries are likely to converge and affect sedimentation zones in a drainage basin as a result of climate change (Macklin et al., 1992; Church et al., 1999), climatic events (Trustrum et al., 1999), major earthquakes (Keefer, 1984), and vegetation changes (Trustrum et al., 1999; Brooks et al., 2003). The wave scale is most relevant to land management because it spans the progress of sediment-related impacts such as inundation, ﬂooding, and effects on aquatic and riparian ecosystems. If so, understanding the dynamics of transport capacity becomes a critical problem for scientists researching the routing of bed material through drainage systems. An entry into investigating relations between sediment transport, supply, and storage is provided by data on degrading, gravel-bed channels from previous ﬁeld studies and experiments (Lisle and Church, 2002). These data show strong, approximately

The evolution of sediment waves in heterogeneous rivers

457

linear relations between transport rate and volume of sediment stored. In this section, I relate more recent investigations.

3.1.

Armoring, transport intensity, and bed elevation

Examples of sediment transport–storage relations for degrading gravel bed channels from ﬂume experiments and two ﬁeld examples (Lisle and Church, 2002) were chosen to neglect variations in channel gradient and to focus on adjustments in bed armoring. In the experiments, clear, uniform ﬂow was run over initially unarmored beds, and sediment output was measured until the bed stabilized. Some experiments showed a marked transition from a condition of weak armoring and high, variable output rates created by migrating bedforms to one of intensifying armoring and declining transport rates until the bed stabilized. (The former state is termed ‘Phase I’ or ‘weakly armored’; the latter is termed ‘Phase II’, or ‘well armored’. Unfortunately, this can create confusion with Phases I and II of Jackson and Beschta (1982), which refer to selective and general transport of sand and gravel under varying discharge.) Other experiments showed only the well armored phase throughout. The transition to increased armoring in these experiments occurred in a range ð0:03oqs o0:16Þ corresponding to some of the highest dimensionless bedload transport rates measured in natural channels, including unarmored desert channels (Reid and Laronne, 1995) and channels carrying high loads of sediment from recent volcanic eruptions (Pitlick, 1992; Hayes et al., 2002). [Dimensionless bedload transport rate is expressed as qs ¼ qs =ðgRD350 Þ1=2 , where qs is the volumetric transport rate per unit width, g is gravitational acceleration, R is submerged speciﬁc gravity of sediment, and D50 is median particle diameter of the bed surface.] Values of qs for each of these and other channels plot as a locus of points along a general relation with dimensionless tractive force or Shields stress, expressed as t50 ¼ t=RD50s , where t is mean boundary shear stress [Fig. 17.3; similar plots presented by Reid and Laronne (1995) and Hayes et al. (2002)]. Data from more mobile, sediment-charged channels plot high along the global trend and show a lower slope in the relation between qs and t50 than do armored channels with lesser loads. The change in slope in these relations appears at a value of qs 0:1 and a value of t50 0:1. In this range of t50 , armoring largely disappears (Parker and Klingeman, 1982; Wilcock and Southard, 1989). The transport formula of Wilcock and Crowe (2003) is non-linear, but shows an increase in curvature at approximately the same value. The agreement of the formula with the data is only fair at high Shields stress. Much of the discrepancy may arise from the tendency for greater abundance of sand on the bed surface at high Shields stress, indicating a shift in appropriate transport relations from sand-poor to sand-rich conditions. Taken together, these observations are consistent with the occurrence of an inﬂection in transport–tractive force relations at a point at which armoring diminishes. Below this transition, a mobile armor inhibits bed mobility and makes transport highly sensitive to variations in tractive force. Above this transition, armoring has little inﬂuence on bed mobility, and transport increases more gradually with applied stress. Weak armoring is commonly associated with aggraded channels (Lisle and Madej, 1992; Gomez et al., 2001; Hayes et al., 2002) and indicates maximum

458

T.E. Lisle

Figure 17.3. Relations between dimensionless bedload transport rate (qs ) and Shield stress (t50 ) for some gravel-bed rivers in humid climates, adapted from Reid and Laronne (1995), Hayes et al. (2002), and Lisle and Smith (2003) and updated by Karen Gran, University of Washington, 2004. Also shown are bedload transport relations of Wilcock and Crowe (2003) for gravel and a gravel–sand mixture of 420% sand. Data are from Milhous (1973), Lisle (1989), Pitlick (1992), Andrews (1994), and Hayes et al. (2002).

transport capacity (Dietrich et al., 1989); strong armoring is associated with stable channels (Andrews, 1984). Increased armoring underlies inﬂections in both the transport–tractive force relation (Fig. 17.3) and transport–storage relations observed in some degrading gravelbed channels (Lisle and Church, 2002). This leads to the following interpretation: In a

The evolution of sediment waves in heterogeneous rivers

459

sediment-charged river where armoring is weak and transport rates are high, transport rates respond weakly to changes in sediment supply, and the channel adjusts primarily by erosion or deposition. At some stage in degradation, armoring strengthens and responds to variations in sediment input rates, causing sediment transfer rates to decline rapidly as supply and storage decline. Given no change in hydrologic regime, a decrease in supply would increase armoring and cause Shields stress at a given discharge to decrease, shifting the locus of points on the transport–tractive stress curve downward. This conceptual model has not been tested in a natural system, but Sawada et al. (1985) and Laronne et al. (2001) report shifts in bedload rating curves with changes in the abundance of gravel and sand on stream beds that are dominated by cobbles and boulders, and Lisle et al. (2000) correlate high reachaveraged, bank-full Shields stress with sediment supply. A change in transport rates corresponding to a change from a well-armored bed to a weakly-armored bed can be estimated from common values of Shields stress for well-armored channels and the Wilcock–Crowe transport function. Andrews (1984) reports an average bank-full Shield stress value of 0.046 for 24 rivers in Colorado, which have low bed-material supplies. If we assume that an average degree of armoring (ratio of median sizes of surface material and bedload) equals four and allow for hydraulic changes due to bed roughness, bank-full bedload transport ranging from a well-armored to weakly-armored condition would vary over approximately 1–2 orders of magnitude. This suggests a need to consider variations in transport capacity in bed-material routing schemes. The weakly armored phase is traditionally identiﬁed as ‘transport-limited’, meaning that sediment-supply rates equal or exceed a ﬁxed value of transport capacity, and ﬂux rates depend solely on discharge. Dietrich et al. (1989) identify transport capacity as being met when armoring diminishes with greater sediment supply. Such a high reference value of transport capacity may be appropriate for landscape evolution models used to predict the construction of alluvial landforms during periods of high sedimentation. But because armoring is prevalent in humidtemperate gravel-bed channels, sediment routing schemes may be improved by incorporating the dynamics of transport capacity under varying supply.

3.2.

Does sediment storage decrease exponentially?

Exponential decay is commonly attributed to time trends in sediment storage and output rates as channels degrade (Graf, 1977; Simon, 1992; Gran and Montgomery, 2005). Although it appears to be a reasonable approximation, the exponential model ﬁts a special case of a ‘well-mixed’ reservoir in which each unit volume of stored sediment has an equal probability of being transferred out of storage, thus the same proportion of the remaining volume is released during each time step (Kirchner et al., 2000). The heterogeneity of natural sediment reservoirs, including channel beds, bars, ﬂoodplains, and other off-channel alluvial features, suggests that the availability and mobility of sediment for transfer is non-uniform (Kelsey et al., 1987). For example, the marginalization of deposits left by migrating channels can render them less likely to be eroded (James, 1991; Nakamura and Kikuchi, 1996). The growth of riparian vegetation may increase the resistance to erosion (e.g., Simon et al., 2004; Millar,

T.E. Lisle

460

2000), and armoring may produce non-linear, negative feedback to progressive scour (Lisle and Church, 2002). If so, a more general model of sediment release is required. One of several possible models is a scaled gamma distribution, expressed here in terms of sediment storage, V ðtÞ ¼ M Z1 GðaÞ ¼

ka ta1 kt e GðaÞ

(17.2)

et ta1 dt

(17.3)

0

where V(t) is the volume of sediment stored in a reservoir at time t, and a and k are constants. The exponential equation is a special case where a ¼ 1. Assuming this value and expressing the original volume store as V0 ¼ M/k, equation (17.1) becomes a simple exponential equation V ðtÞ ¼ V 0 ekt The exponential equation can conform to time trends in sediment evacuation in natural channels, given uncertainties in unsteady ﬂow effects and the value of the datum to determine V0 (Lisle and Church, 2002). Nevertheless, a sediment reservoir with resistant marginal stores or non-linear effects of developing armoring may be modeled more accurately by a scaled gamma distribution with ao1 (Fig. 17.4). Therefore, allowing a to vary offers a more general solution to time trends in sediment storage. Gamma distributions with ao1 would manifest an initial rapid depletion and a long elevated tail of slowly decreasing residual stores as more resistant or marginalized storage components are eroded. For watershed managers, such a model implies prolonged effects from sedimentary disturbances. 3.3.

Transport– storage relations during full episodes of aggradation and degradation

Although it is well known that sediment inputs can increase transport rates and cause aggradation, relations between transport and storage as bed elevations rise and fall in response to changes in sediment supply are poorly understood. Smith (2004) conducted a ﬂume experiment to investigate relations between sediment transport and storage during episodes of aggradation and degradation in Cuneo Creek in northern California. Much of the aggradation occurred during extreme Figure 17.4. Examples of time trends in sediment transfer rates. Transfer rates are shown instead of storage volumes to avoid the issue of the uncertainty of the datum from which to measure storage volume. (The derivative of an exponential relation, e.g., as the transfer rate is to the stored volume, is also exponential.) In each plot, a line describes the best-ﬁt exponential relation. (A) Higashi-Gouchi is a steep channel in the Japanese Alps that temporarily stored debris ﬂow deposits in 1982–1986 (Maita, 1991). Time steps are intervals between ﬂood events. (B) Experimental run 3.4 of Little and Mayer (1972) began with a screeded bed under uniform ﬂow and armored as the bed degraded. The pattern in output rate shows a strong departure from an exponential model. (C) Experimental run 1.7 by Profﬁtt (1980) was conducted similarly to that of Little and Mayer but shows a closer ﬁt to an exponential model.

The evolution of sediment waves in heterogeneous rivers

461

T.E. Lisle

462 Table 17.3.

Experimental conditions.

Parameter

Value

Flume length, width Slope Discharge Scale factor Sediment distribution

12 m, 0.76 m 3% 1.1 l/s 25 Range ¼ 0.25–11.2 mm Median ¼ 1 mm Standard Deviation ¼ 1.8 mm

ﬂoods that caused massive landsliding and gullying in the disturbed basin. High supply rates were not maintained following these events and the channel soon began to degrade. Consequently, equilibrium between sediment input and output were never achieved system wide. Here, I present results from one of the experimental runs modeling an episode of aggradation and degradation (Table 17.3). In Run 2, sediment of the same mixture as that of the bed was fed at a low bulk rate of 5.9 cm3/s until equilibrium was achieved. Feed rate was then increased to 10.9 cm3/s and thereafter decreased in two steps – ﬁrst to the original rate (5.9 cm3/s) and then to 3.1 cm3/s. Equilibrium was approximately restored in each step of degradation. Sediment output rate was measured continuously, and the ﬂume was shut down periodically to measure bed topography with high-resolution laser scanning. Initially, the channel was well armored and contained a single-thread, alternate bar sequence. When feed rate was increased, the channel became covered with mobilized particles and armoring diminished. While ﬁne particles blanketed most of the bed, small congested zones of large particles formed and broke up. The active bed widened, and the channel became braided. During aggradation, output rates increased gradually but were never more than about one-half of the feed rate (Fig. 17.5A). When the feed rate was reduced, the feed rate and output rate became equal for a short time. Hysteresis loops in the relation between output rate and sediment storage described the exit of small sediment pulses. Afterward, a single-thread channel incised, and wellspaced granules and pebbles traveled rapidly over a bed of mobile sand. The channel appeared to be highly efﬁcient at transporting bedload, and this was conﬁrmed by high output rates. Later, the channel widened and then armored. The zone of active transport contracted and output rates decreased to equal the original feed rate. The zone contracted further after the second reduction in feed rate. The relation between transport and storage during the sequence of aggradation and degradation in Run 2 was more complex than those that have been observed from previous experiments that involved degradation only (Lisle and Church, 2002). Transport rates during aggradation were lower than peak rates during the unarmored phase of degradation, creating strong counter-clockwise hysteresis in transport–storage relations. After degradation, the bed stabilized at a higher elevation than had existed before the cycle of aggradation and degradation. The increase in apparent base level was due to the accumulation of large particles during each phase of degradation.

The evolution of sediment waves in heterogeneous rivers

463

Figure 17.5. Patterns of variation of sediment output rate and stored volume from Run 2 (Smith, 2004). Fig. A represents the entire 12 m ﬂume; Fig. B represents a 1 m section from 6.3–7.3 m.

Some hysteresis is an inevitable result of the lag in output to changes in input at the head of a ﬂume. The amplitude of hysteresis can be anticipated from application of the Exner equation to a transport–storage plot. A channel of a given length that responds to sediment input with rapid deposition (i.e., high @z=@t) relative to transport rate has pronounced streamwise variation in transport rate (i.e., high @qs =@x), a strong lag in sediment transfer and high-amplitude hysteresis. Conversely, a channel of the same length that responds to an input with an increase in transport rate and little storage exhibits less pronounced hysteresis. To remove the lag effect analytically, transfer rate and storage volume in short (1m) sections were computed from stepwise budgets between the input and outlet of the ﬂume for periods between the occasional topographic surveys. Relations for one section near the middle of the ﬂume are shown in Fig. 17.5B. Much of the lag-related hysteresis is no longer evident, but changes in the relation between periods of aggradation and degradation remain. Starting with the aggradation phase, transfer rates varied less with storage as the channel widened and became braided than during the later degradation phase when the channel underwent a large change in morphology and texture. Although high transport rates are commonly associated with braiding, clusters of large particles formed brieﬂy over most of the bed and prevented the integration of smooth zones (Iseya and Ikeda, 1987) that could efﬁciently transfer all sediment sizes. After the feed rate was reduced, many large particles came to rest, allowing the abundant ﬁne particles on the bed to become winnowed and concentrated in extensive smooth zones. Sheets of material of all sizes migrated down the channel. Peak transport rates for Run 2 were achieved when a single channel became better deﬁned and began to incise. Incision increased ﬂow depth, thereby increasing

T.E. Lisle

464

boundary shear stress acting on a zone of full transport that spanned most of the deﬁned channel. Later, the zone of transport contracted and armoring increased as transport rates decreased. These observations indicate that not all of the hysteresis in transport–storage relations for the ﬂume as a whole was due to the lag in output behind changes in input, but can be partly attributed to changes in channel morphology and texture. These results indicate that sediment reservoirs do not behave entirely as entities governed by characteristics at the reservoir scale, but instead adjust to changes in base level and sediment supply both internally and between adjacent reservoirs. Transport capacity is apparently conditioned by changes in local sediment supply in response to larger scale changes such as those created by sediment waves (Lane et al., 1996). Proper scaling is important in accurately representing the dynamics of transport– storage relations in sediment routing models. A coarse-scale routing model based on sediment transfers through series of sediment reservoirs may perform best for long time scales (100–102 yr) in stepped proﬁles where lower bounds of high-storage reservoirs are marked by relatively stable reaches with steep transport-storage relations.

4.

Conclusions

Relative tendencies for sediment waves to disperse or translate have served as a discussion point on overall wave behavior, but this issue is diminished as new numerical models promise detailed solutions to more practical problems. Translation does not appear to be an outcome of the interaction of ﬂow, sediment transport, and the proﬁle of a sediment wave under hydraulic conditions typical of gravel-bed rivers. However, large sediment inputs are rarely of the same particle-size distribution as the ambient bed material, and selective transport can greatly affect wave evolution. Coarse inputs can cause waves to grow as they incorporate ﬁner ambient material and reduce its mobility. Fine inputs can promote translation in hydraulic environments (high Froude number) where waves would otherwise only disperse. Even with this effect, dispersion appears to be the dominant if not exclusive trait of the evolution of mega-slugs and super-slugs. But in the ﬁnal analysis, non-uniformity in natural channels commonly creates streamwise variations in transport capacity that superimpose patterns in deposition and erosion on inherent behaviors of sediment waves. Comprehensive ﬁeld studies are motivated by the complexity of the problem and the limitations of scale for analytically and physically modeling sediment disturbances and the need to investigate ecological consequences of sediment waves in natural channels. One-dimensional models are providing predictions that need to be tested, and multi-dimensional models will need to be grounded in new theory and ﬁeld observations. Existing ﬁeld studies provide valuable contrasts in general behavior, but interpretations are often limited by the paucity of pre-disturbance data on channel conditions, and in many cases, the event that produces the wave also deforms it into advanced stages of evolution. New well-documented case studies involving more combinations of particle size, abrasion rates, and hydraulic conditions are needed. In particular, it would be valuable to test predictions that coarse inputs would accumulate mobile bed material and grow, that high abrasion rates of inputs would

The evolution of sediment waves in heterogeneous rivers

465

promote upstream propagation of waves, and that ﬁne inputs could substantially mobilize and scour bed material downstream. Understanding effects of sediment inputs on the movement of bed material in rivers has advanced in the last decade, but resulting theoretical and numerical models are limited to one dimension. Although these can provide useful general predictions and analytical references for variations in deposition and transfer of sediment downstream, they fall short of solving many of the problems of sediment impacts, which tend to be fully dimensional. Moreover, the redistribution of sediment involves three-dimensional processes including bar instability and exchanges of sediment between channels and ﬂoodplains. Using numerical models to predict channel changes in three dimensions over reaches of length Z103 channel widths would demand prodigious computer memory. Alternatively, one-dimensional sediment-routing models could be coupled with reach-speciﬁc models that predict changes in channel form from changes in sediment ﬂux and storage. This motivates investigations of responses in channel conditions, including morphology and texture, from local changes in supply that govern rates of deposition and transfer of sediment as bedload disturbances propagate downstream. The release of sediment from decommissioned dams presents complete, though non-repeatable experiments on wave evolution and valuable applications of predictive models. Such experiments have been performed (Pizzuto, 2002; Doyle et al., 2003), but there has yet to be a study of the release of large volumes of sediment from a decommissioned dam on a gravel-bed river. The mediator between storage and transfer of sediment pulses is transport capacity. A safe assumption is that over time scales in which sediment disturbances are propagated through a channel system, sediment is transferred not according to a ﬁxed relation with ﬂow, but to some dynamic relation that is governed by changes in channel morphology and texture regulating rates and patterns of sediment transport and storage. Transport and storage are linked indirectly through channel response to streamwise variations in sediment ﬂux resulting from disturbances in hydrologic and sediment regimes. Transport–storage relations are poorly constrained and can be expected to vary between sediment reservoirs delineated by quasi-uniform channels and ﬂoodplains. Furthermore, recent experiments suggest that transport–storage relations for a sediment reservoir can also vary during and between episodes of aggradation and degradation. Storage volumes are the culmination of past changes in supply and hydrologic events mediated by a dynamical adjustment of transport capacity. Finally, transport–storage relations can vary within the reservoir due to lags in the propagation of the pulse through the reservoir and to the associated channel changes at a ﬁner scale. The pursuit of transport–storage relations can increase our understanding of sediment routing processes and the effects of sediment disturbances on aquatic and riparian ecosystems. Someday this may result in an accurate coarse-scale model fashioned around transport–storage relations in a network of reach-scale sediment reservoirs.

Acknowledgments The paper was improved by reviews by Marwan Hassan, Jonathan Laronne, Gary Parker, and Bonnie Smith.

466

T.E. Lisle

References Andrews, E.D., 1984. Bed-material entrainment and hydraulic geometry of gravel-bed rivers in Colorado. Geol. Soc. Am. Bull. 95, 371–378. Andrews, E.D., 1994. Marginal bed load transport in a gravel bed stream, Sagehen Creek, California. Water Resour. Res. 30, 2241–2250. Ashworth, P.J., 1996. Mid-channel bar growth and its relationship to local ﬂow strength and direction. Earth Surf. Process. Landf. 21, 103–123. Bartley, R., Rutherford, I.D., 2005. Re-evaluation of the wave model as a tool for quantifying the geomorphic recovery potential of streams disturbed by sediment slugs. Geomorphology 64, 221–242. Benda, L.E., Dunne, T., 1997. Stochastic forcing of sediment routing and storage in channel networks. Water Resour. Res. 33, 2865–2880. Benda, L.E., Miller, D.J., Bigelow, P., Andras, K., 2004. Effects of post-wildlife erosion on channel environments Boise River, Idaho. Forest Ecol. Manag. 178, 105–119. Beschta, R.L., 1983. Channel changes following storm-induced hillslope erosion in the Upper Kowai Basin, Torlesse Range, New Zealand. J. Hydrol. (NZ) 22, 93–111. Brooks, A.P., Brierley, G.J., Millar, R.G., 2003. The long-term control of vegetation and woody debris on channel and ﬂood-plain evolution: insights from a paired catchment study in southeastern Australia. Geomorphology 51, 7–30. Brummer, C.J., Montgomery, D.R., 2006. Inﬂuence of coarse lag formation on the mechanics of sediment pulse dispersion in a mountain stream, Squire Creek, North Cascades, Washington, United States. Water Resour. Res. 42. Cao, Z., Carling, P.A., 2003. On evolution of bed material waves in alluvial rivers. Earth Surf. Process. Landf. 28, 437–441. Cao, Z., Carling, P.A., 2005. Further perspectives on the evolution of bed material waves in alluvial rivers. Earth Surf. Process. Landf. 30, 115–120. Church, M., 1983. Pattern of instability in a wandering gravel bed channel. In: Collinson, J.D. and Lewin, J. (Eds), Modern and Ancient Fluvial Systems. Blackwell Scientiﬁc Publications, Oxford, pp. 169–180. Church, M., Ham, E., Hassan, M., Slaymaker, O., 1999. Fluvial clastic sediment yield in Canada: scaled analysis. Can. J. Earth Sci. 36, 1267–1280. Church, M., Jones, D., 1982. Channel bars in gravel-bed rivers. In: Hey, R.D., Bathurst, J.C., and Thorne, C.R. (Eds), Gravel-Bed Rivers. Wiley, Chichester, pp. 291–338. Cui, Y., Parker, G., 2005. Numerical model of sediment pulses and sediment supply disturbances in mountain rivers. J. Hydraul. Eng. 131, 646–656. Cui, Y., Parker, G., Braudrick, C.A., et al., 2006. Dam removal express assessment models (DREAM). Part 1. Model development and validation. J. Hydraul. Res. 44, 291–307. Cui, Y., Parker, G., Lisle, T.E., et al., 2003a. Sediment pulses in mountain rivers. Part 1. Experiments. Water Resour. Res. 39, 1239, doi:1210.1029/2002WR001803. Cui, Y., Parker, G., Lisle, T.E., et al., 2005. More on the evolution of bed material waves in alluvial rivers. Earth Surf. Process. Landf. 30, 107–114. Cui, Y., Parker, G., Pizzuto, J.E., Lisle, T.E., 2003b. Sediment pulses in mountain rivers. Part 2. Comparison between experiments and numerical predictions. Water Resour. Res. 39, 1240, doi:1210.1029/ 2002WR001805. Cui, Y., Wilcox, A.C., 2005. Numerical modeling of sediment transport upon dam removal: application to Marmot Dam in Sandy River, Oregon. In: Garcia, M.H. (Ed.), Sedimentation Engineering, ASCE Manual 54. American Society of Civil Engineers, Reston, VA. Dietrich, W.E., Kirchner, J.W., Ikeda, H., Iseya, F., 1989. Sediment supply and the development of the coarse surface layer in gravel-bedded rivers. Nature 340, 215–217. Dietrich, W.E., Smith, J.D., 1983. Inﬂuence of the point bar on ﬂow through curved channels. Water Resour. Res. 19, 1173–1192. Dodd, A.M., 1998. Modeling the Movement of Sediment Waves in a Channel. M. S. Humboldt State University, Arcata.

The evolution of sediment waves in heterogeneous rivers

467

Doyle, M.W., Stanley, E.H., Harbor, J.M., 2003. Channel adjustments following two dam removals in Wisconsin. Water Resour. Res. 39, 1011, doi:1010.1029/2002WR001714. Gilbert, G.K. (1917) Hydraulic mining debris in the Sierra Nevada. US Geological Survey Professional Paper 105, Department of the Interior, 154pp. Gomez, B., Rosser, B.J., Peacock, D.H., et al., 2001. Downstream ﬁning in a rapidly aggrading gravel bed river. Water Resour. Res. 37, 1813–1823. Gott, J. (1998) The behavior of ﬁne debris-torrent derived sediments in the North Fork Boise River, 1995–1997 (abstract). In: Western Watersheds, 7th Biennial Watershed Management Council Conference. Watershed Management Council, Boise, ID. Graf, W.L., 1977. The rate law in ﬂuvial geomorphology. Am. J. Sci. 277, 178–191. Gran, K.B., Montgomery, D.R., 2005. Spatial and temporal patterns in ﬂuvial recovery following volcanic eruptions: Channel response to basin-wide sediment loading at Mount Pinatubo, Phillipines. Geol. Soc. Am. Bull. 117, 195–211. Hassan, M. et al., this volume. Sediment storage and transport in coarse bed streams: scale condsiderations. Hayes, S.K., Montgomery, D.R., Newhall, C.G., 2002. Fluvial sediment transport and deposition following the 1991 eruption of Mount Pinatubo. Geomorphology 45, 211–224. Hoey, T., 1992. Temporal variations in bedload transport rates and sediment storage in gravel-bed rivers. Prog. Phys. Geogr. 16, 319–338. Hooke, J., 2003. Coarse sediment connectivity in river channel systems: a conceptual framework and methodology. Geomorphology 56, 79–94. Iseya, F., Ikeda, H., 1987. Pulsations in bedload transport rates induced by a longitudinal sediment sorting: a ﬂume study using sand and gravel mixtures. Geogr. Ann. 69A, 15–27. Jackson, R.G., 1975. Hierarchical attributes and a unifying model of bedforms composed of cohesionless material and produced by shearing ﬂow. Geol. Soc. Am. Bull. 86, 1523–1533. Jackson, W.L., Beschta, R.L., 1982. A model of two-phase bedload transport in an Oregon Coast Range stream. Earth Surf. Process. Landf. 7, 517–527. James, L.A., 1991. Incision and morphologic evolution of an alluvial channel recovering from hydraulic mining sediment. Geol. Soc. Am. Bull. 91, 723–736. James, L.A., 2006. Bed waves at the basin scale: implications for river management and restoration. Earth Surf. Process. Landf. 31, 1692–1706. Jones, L.S., Humphrey, N.F., 1997. Weathering-controlled abrasion in a coarse-grained, meandering reach of the Rio Grande; implications for the rock record. Geol. Soc. Am. Bull. 109, 080–1088. Kasai, M., Marutani, T., Brierley, G.J., 2004. Patterns of sediment slug translation and dispersion following typhoon-induced disturbance, Oyabu Creek, Kyushu, Japan. Earth Surf. Process. Landf. 29, 59–76. Keefer, D.K., 1984. Landslides caused by earthquakes. Geol. Soc. Am. Bull. 95, 406–421. Kelsey, H.M., Lamberson, R.L., Madej, M.A., 1987. Stochastic model for the long-term transport of stored sediment in a river channel. Water Resour. Res. 23, 1738–1750. Kirchner, J.W., Fong, X., Neal, C., 2000. Fractal stream chemistry and its implications for contaminant transport in catchments. Nature 403, 524–527. Knighton, A.D., 1989. River adjustments to changes in sediment load: the effects of tin mining on the Ringarooma River, Tasmania, 1875–1984. Earth Surf. Process. Landf. 14, 333–359. Korup, O., 2004. Landslide-induced river channel avulsions in mountain catchments of southwest New Zealand. Geomorphology 63, 57–80. Lane, S.N., Richards, K.S., Chandler, J.H., 1996. Discharge and sediment supply controls on erosion and deposition in a dynamic alluvial channel. Geomorphology 15, 1–15. Laronne, J.B., Garcia, C., Reid, I., 2001. Mobility of patch sediment in gravel bed streams: patch character and its implications for bedload. In: Mosley, M.P. (Ed.), Gravel-Bed Rivers V. New Zealand Hydrological Society, Wellington, New Zealand, pp. 249–280. Lisle, T.E., 1989. Sediment transport and resulting deposition in spawning gravels, north coastal California. Water Resour. Res. 25, 1303–1319. Lisle, T.E., Church, M., 2002. Sediment transport-storage relations for degrading gravel-bed channels. Water Resour. Res. 38, 1219, doi:1210.1029/2001WR001086.

468

T.E. Lisle

Lisle, T.E., Cui, Y., Parker, G., et al., 2001. The dominance of dispersion in the evolution of bed material waves in gravel bed rivers. Earth Surf. Process. Landf. 26, 1409–1420. Lisle, T.E., Madej, M.A., 1992. Spatial variation in armouring in a channel with high sediment supply. In: Billi, R.D.H.P., Thorne, C.R., and Tacconi, P. (Eds), Dynamics of Gravel-bed Rivers. John Wiley, Chichester, pp. 277–291. Lisle, T.E., Nelson, J.M., Pitlick, J., et al., 2000. Variability of bed mobility in natural gravel-bed channels and adjustments to sediment load at the local and reach scales. Water Resour. Res. 36, 3743–3756. Lisle, T.E., Pizzuto, J.E., Ikeda, H., et al., 1997. Evolution of a sediment wave in an experimental channel. Water Resour. Res. 33, 1971–1981. Lisle, T.E. and Smith, B., 2003. Dynamic transport capacity in gravel-bed river systems. In: Araya, T., Kuroki, M. and Marutani, T. (Eds), International Workshop for Source to Sink Sedimentary Dynamics in Catchment Scale. Organizing Committee of the International Workshop for Sedimentary Dynamics, Sapporo, Japan, pp. 187–206. Little, W.C., Mayer, P.G., 1972. The role of sediment gradation of channel armoring. ERC-0672. Georgia Institute of Technology, Atlanta. Lyons, J.K., Beschta, R.L., 1983. Land use, ﬂoods, and channel changes: upper middle Fork Willamette River, Oregon (1936–1980). Water Resour. Res. 19, 463–471. Macklin, M.G., Rumsby, B.T., Heap, T., 1992. Flood alluviation and entrenchment: Holocene valley-ﬂoor development and transformation in the British uplands. Geol. Soc. Am. Bull. 104, 631–643. Madej, M.A., Ozaki, V., 1996. Channel response to sediment wave propagation and movement, Redwood Creek, California, USA. Earth Surf. Process. Landf. 21, 911–927. Maita, H., 1991. Sediment dynamics of a high gradient stream in the Oi River Basin of Japan. Gen. Technical Report PSW-GTR-130, USDA Forest Service, Montreal, Canada. Meade, R.H., 1982. Sources, sinks, and storage of river sediment in the Atlantic drainage of the United States. J. Geol. 90, 235–252. Meade, R.H., 1985. Wavelike movement of bedload sediment, East Fork River, Wyoming. Environ. Geol. Water Sci. 7, 215–225. Milhous, R.T., 1973. Sediment transport in a gravel-bottomed stream, PhD. Thesis, Oregon State University, Corvallis, OR. Miller, D.J., Benda, L.E., 2000. Effects of punctuated sediment supply on valley-ﬂoor landforms and sediment transport. Geol. Soc. Am. Bull. 112, 1814–1824. Millar, R.G., 2000. Inﬂuence of bank vegetation on alluvial channel patterns. Water Resour. Res. 36, 1109–1118. Nakamura, F., Kikuchi, S., 1996. Some methodological developments in the analysis of sediment transport processes using age distribution of ﬂoodplain deposits. Geomorphology 16, 139–145. Nakamura, F., Maita, H., Araya, T., 1995. Sediment routing analysis based on chronological changes in hillslope and riverbed morphologies. Earth Surf. Process. Landf. 20, 333–346. Neill, C.R., 1987. Sediment balance considerations linking long-term transport and channel processes. In: Thorne, C.R., Bathurst, J.C., and Hey, R.D. (Eds), Sediment Transport in Gravel-Bed Rivers. John Wiley, Chichester, pp. 225–240. Nicholas, A.P., Ashworth, P.J., Kirkby, M.G., et al., 1995. Sediment slugs: large-scale ﬂuctuations in ﬂuvial sediment transport rates and storage volumes. Prog. Phys. Geogr. 19, 500–519. Parker, G., Klingeman, P.C., 1982. On why gravel bed streams are paved. Water Resour. Res. 18, 1409–1423. Pickup, G., Higgins, R.J., 1979. Estimating sediment transport in a braided gravel channel – The Kawerong River, Bougainville, Papua New Guinea. J. Hydrol. 40, 283–297. Pickup, G., Higgins, R.J., Grant, I., 1983. Modelling sediment transport as a moving wave – the transfer and deposition of mining waste. J. Hydrol. 60, 281–301. Pitlick, J., 1992. Flow resistance under conditions of intense gravel transport. Water Resour. Res. 28, 891–903. Pitlick, J., 1993. Response and recovery of a subalpine stream following a catastrophic ﬂood. Geol. Soc. Am. Bull. 105, 657–670.

The evolution of sediment waves in heterogeneous rivers

469

Pizzuto, J.E., 2002. Effects of dam removal on river form and process. BioScience 52, 683–692. Profﬁtt, G.T., 1980. Selective transport and armouring of non-uniform alluvial sediments. 80/22, Department of Civil Engineering, University of Canterbury, Christchurch, NZ. Reid, I., Frostick, L.E., Layman, J.T., 1985. The incidence and nature of bedload transport during ﬂood ﬂows in coarse-grained alluvial channels. Earth Surf. Process. Landf. 10, 33–44. Reid, I., Laronne, J.B., 1995. Bed load sediment transport in an ephemeral stream and a comparison with seasonal and perennial counterparts. Water Resour. Res. 31, 773–781. Sawada, T., Ashida, K., and Takahashi, T., 1985. Sediment transport in mountain basins. In: International Symposium on Erosion, Debris Flow and Disaster Prevention, Tsukuba, Japan. pp. 139–144. Simon, A., 1992. Energy, time, and channel evolution in catastrophically disturbed ﬂuvial systems. Geomorphology 5, 345–372. Simon, A., Bennett, S.J., Neary, V.S., 2004. Riparian vegetation and geomorphology: problems and opportunities. In: Bennett, S.J. and Simon, A. (Eds), Riparian Vegetation and Fluvial Geomorphology. American Geophysical Union, Washington, DC, pp. 1–10. Smith, B.J., 2004. Relations between bed material transport and storage during aggradation and degradation in a gravel bed channel. Masters thesis. Humboldt State University, Arcata, CA. Sternberg, H., 1875. Untersuchungen ueber laengen- und querproﬁl geschiebefuehrende ﬂusse. Z. Bauwesen 25, 483–506. Sutherland, D.G., Hansler, M.E., Hilton, S., Lisle, T.E., 2002. Evolution of a landslide-induced sediment wave in the Navarro River, California. Geol. Soc. Am. Bull. 114, 1036–1048. Trimble, S.W., 1981. Changes in sediment storage in the Coon Creek Basin, Driftless Area, Wisconsin, 1853 to 1975. Science 214, 181–183. Trustrum, N.A., Gomez, B., Page, M.J., et al., 1999. Sediment production, storage and output: the relative role of large magnitude events in steepland catchments. Z. Geomorphol. Suppl. 115, 71–86. Turner, T.R., 1995. Geomorphic response of the Madison River to point sediment loading at the Madison Slide, southwest Montana. M. S. Montana State University. Wathen, S.J., Hoey, T.B., 1998. Morphologic controls on the downstream passage of a sediment wave in a gravel-bed stream. Earth Surf. Process. Landf. 23, 715–730. Werritty, A., 1992. Downstream ﬁning in a gravel-bed river in southern Portland: lithologic controls and the role of abrasion. In: Billi, P., Hey, R.D., Thorne, C.R., and Tacconi, P. (Eds), Dynamics of Gravel-Bed Rivers. Wiley, Chichester, pp. 333–350. Wilcock, P.R., Crowe, J.C., 2003. Surface-based transport model for mixed-size sediment. J. Hydraul. Eng. 129, 120–127. Wilcock, P.R., Southard, J.B., 1989. Bed load transport of mixed size sediment: fractional transport rates, bed forms, and the development of a coarse bed surface layer. Water Resour. Res. 25, 1629–1641. Wohl, E., Cenderelli, D.A., 2000. Sediment deposition and transport patterns following a reservoir sediment release. Water Resour. Res. 36, 319–333.

Discussion by Rob Ferguson A distinction is often made between ‘transport-limited’ and ‘supply-limited’ bedload ﬂux. I would welcome your opinion on the utility of this distinction in the light of your excellent review of sediment pulses. Your case studies show that a sudden injection of sediment into a gravel-bed river leads to a period of mutual adjustment amongst bed level, surface grain size distribution (GSD), and bedload ﬂux. There is a shifting relationship between bedload ﬂux and water discharge at any one location, but it seems to me that the immediate reason for this is the change in surface GSD; supply to the system is only indirectly relevant through the history of change in bed level and surface GSD. Would ‘availability-limited’ be a better term?

The evolution of sediment waves in heterogeneous rivers

469

Pizzuto, J.E., 2002. Effects of dam removal on river form and process. BioScience 52, 683–692. Profﬁtt, G.T., 1980. Selective transport and armouring of non-uniform alluvial sediments. 80/22, Department of Civil Engineering, University of Canterbury, Christchurch, NZ. Reid, I., Frostick, L.E., Layman, J.T., 1985. The incidence and nature of bedload transport during ﬂood ﬂows in coarse-grained alluvial channels. Earth Surf. Process. Landf. 10, 33–44. Reid, I., Laronne, J.B., 1995. Bed load sediment transport in an ephemeral stream and a comparison with seasonal and perennial counterparts. Water Resour. Res. 31, 773–781. Sawada, T., Ashida, K., and Takahashi, T., 1985. Sediment transport in mountain basins. In: International Symposium on Erosion, Debris Flow and Disaster Prevention, Tsukuba, Japan. pp. 139–144. Simon, A., 1992. Energy, time, and channel evolution in catastrophically disturbed ﬂuvial systems. Geomorphology 5, 345–372. Simon, A., Bennett, S.J., Neary, V.S., 2004. Riparian vegetation and geomorphology: problems and opportunities. In: Bennett, S.J. and Simon, A. (Eds), Riparian Vegetation and Fluvial Geomorphology. American Geophysical Union, Washington, DC, pp. 1–10. Smith, B.J., 2004. Relations between bed material transport and storage during aggradation and degradation in a gravel bed channel. Masters thesis. Humboldt State University, Arcata, CA. Sternberg, H., 1875. Untersuchungen ueber laengen- und querproﬁl geschiebefuehrende ﬂusse. Z. Bauwesen 25, 483–506. Sutherland, D.G., Hansler, M.E., Hilton, S., Lisle, T.E., 2002. Evolution of a landslide-induced sediment wave in the Navarro River, California. Geol. Soc. Am. Bull. 114, 1036–1048. Trimble, S.W., 1981. Changes in sediment storage in the Coon Creek Basin, Driftless Area, Wisconsin, 1853 to 1975. Science 214, 181–183. Trustrum, N.A., Gomez, B., Page, M.J., et al., 1999. Sediment production, storage and output: the relative role of large magnitude events in steepland catchments. Z. Geomorphol. Suppl. 115, 71–86. Turner, T.R., 1995. Geomorphic response of the Madison River to point sediment loading at the Madison Slide, southwest Montana. M. S. Montana State University. Wathen, S.J., Hoey, T.B., 1998. Morphologic controls on the downstream passage of a sediment wave in a gravel-bed stream. Earth Surf. Process. Landf. 23, 715–730. Werritty, A., 1992. Downstream ﬁning in a gravel-bed river in southern Portland: lithologic controls and the role of abrasion. In: Billi, P., Hey, R.D., Thorne, C.R., and Tacconi, P. (Eds), Dynamics of Gravel-Bed Rivers. Wiley, Chichester, pp. 333–350. Wilcock, P.R., Crowe, J.C., 2003. Surface-based transport model for mixed-size sediment. J. Hydraul. Eng. 129, 120–127. Wilcock, P.R., Southard, J.B., 1989. Bed load transport of mixed size sediment: fractional transport rates, bed forms, and the development of a coarse bed surface layer. Water Resour. Res. 25, 1629–1641. Wohl, E., Cenderelli, D.A., 2000. Sediment deposition and transport patterns following a reservoir sediment release. Water Resour. Res. 36, 319–333.

Discussion by Rob Ferguson A distinction is often made between ‘transport-limited’ and ‘supply-limited’ bedload ﬂux. I would welcome your opinion on the utility of this distinction in the light of your excellent review of sediment pulses. Your case studies show that a sudden injection of sediment into a gravel-bed river leads to a period of mutual adjustment amongst bed level, surface grain size distribution (GSD), and bedload ﬂux. There is a shifting relationship between bedload ﬂux and water discharge at any one location, but it seems to me that the immediate reason for this is the change in surface GSD; supply to the system is only indirectly relevant through the history of change in bed level and surface GSD. Would ‘availability-limited’ be a better term?

470

T.E. Lisle

Discussion by Ian Reid Lisle’s excellent synthesis of the impact of sediment waves on sediment supply and the non-stationarity of at-a-station transport capacity helps us further contextualise the enormous scatter in the bedload-shear stress relation exhibited by many armored gravel-bed channels (e.g., see Reid and Laronne, 1995). For the purpose of predicting bedload in this kind of channel, it reﬂects the need for a suite of rating curves that relate bedload ﬂux to hydraulics, some of which will be appropriate when the passage of a wave dictates limitation on sediment supply, whilst others will sufﬁce to cover periods when supply is less limited. However, the irregular and episodic incidence of wave generation shown by Lisle for streams of the North American West coast rivers (as distinct, perhaps, from those of snow-melt regimes such as the East Fork River, Meade et al., 1981) makes the choice of an appropriate rating curve for any particular ﬂood uncertain (a problem also evinced by Laronne et al., 2001 for steep mountain streams). The irregular episodicity of wave formation (reliant as it is on substantial and rapid injections of material from, e.g., hillslope mass failures or effective tributary ﬂoods) also poses problems for other types of predictive sediment routing tools, such as the cellular model being developed by Coulthard et al. (this volume).

References Laronne, J.B., Garcia, C., Reid, I., 2001. Mobility of patch sediment in gravel bed streams: patch character and its implications for bedload. In: Mosley, M.P. (Ed.), Gravel-Bed Rivers V. New Zealand Hydrological Society Inc., Wellington, New Zealand, pp. 249–289. Meade, R.H., Emmett, W.W., and Myrick, R.M. 1981. Movement and storage of bed material during 1979 in East Fork River, Wyoming, USA. In: Erosion and sediment transport in Paciﬁc rim steep lands. International Association of Hydrological Sciences Publication. 132, 225–235. Reid, I., Laronne, J.B., 1995. Bedload sediment transport in an ephemeral stream and a comparison with seasonal and perennial counterparts. Water Resour. Res. 31, 773–781.

Reply by the author I thank Rob Ferguson and Ian Reid for thoughtful discussions that lead logically to some important problems left unresolved in my paper. By questioning the utility of the terms, ‘transport-limited’ and ‘supply-limited’ in bedload regimes, Ferguson helps to widen the crack in commonly used idealizations that has limited perspectives on the range of channel adjustments to a continuum of variations in sediment supply. These terms are end members that could realistically be represented by concrete raceways and terminal deltas, but not by most alluvial channels. Our premise in Lisle and Church (2002) is that all alluvial channels respond to a change in sediment supply with a change in both transport rate and channel storage, the relative degree of each depending on background conditions and geomorphic history. A change in bedload transport rating curves for channels with mixed-size loads is commonly associated with a change in armoring and organization

470

T.E. Lisle

Discussion by Ian Reid Lisle’s excellent synthesis of the impact of sediment waves on sediment supply and the non-stationarity of at-a-station transport capacity helps us further contextualise the enormous scatter in the bedload-shear stress relation exhibited by many armored gravel-bed channels (e.g., see Reid and Laronne, 1995). For the purpose of predicting bedload in this kind of channel, it reﬂects the need for a suite of rating curves that relate bedload ﬂux to hydraulics, some of which will be appropriate when the passage of a wave dictates limitation on sediment supply, whilst others will sufﬁce to cover periods when supply is less limited. However, the irregular and episodic incidence of wave generation shown by Lisle for streams of the North American West coast rivers (as distinct, perhaps, from those of snow-melt regimes such as the East Fork River, Meade et al., 1981) makes the choice of an appropriate rating curve for any particular ﬂood uncertain (a problem also evinced by Laronne et al., 2001 for steep mountain streams). The irregular episodicity of wave formation (reliant as it is on substantial and rapid injections of material from, e.g., hillslope mass failures or effective tributary ﬂoods) also poses problems for other types of predictive sediment routing tools, such as the cellular model being developed by Coulthard et al. (this volume).

References Laronne, J.B., Garcia, C., Reid, I., 2001. Mobility of patch sediment in gravel bed streams: patch character and its implications for bedload. In: Mosley, M.P. (Ed.), Gravel-Bed Rivers V. New Zealand Hydrological Society Inc., Wellington, New Zealand, pp. 249–289. Meade, R.H., Emmett, W.W., and Myrick, R.M. 1981. Movement and storage of bed material during 1979 in East Fork River, Wyoming, USA. In: Erosion and sediment transport in Paciﬁc rim steep lands. International Association of Hydrological Sciences Publication. 132, 225–235. Reid, I., Laronne, J.B., 1995. Bedload sediment transport in an ephemeral stream and a comparison with seasonal and perennial counterparts. Water Resour. Res. 31, 773–781.

Reply by the author I thank Rob Ferguson and Ian Reid for thoughtful discussions that lead logically to some important problems left unresolved in my paper. By questioning the utility of the terms, ‘transport-limited’ and ‘supply-limited’ in bedload regimes, Ferguson helps to widen the crack in commonly used idealizations that has limited perspectives on the range of channel adjustments to a continuum of variations in sediment supply. These terms are end members that could realistically be represented by concrete raceways and terminal deltas, but not by most alluvial channels. Our premise in Lisle and Church (2002) is that all alluvial channels respond to a change in sediment supply with a change in both transport rate and channel storage, the relative degree of each depending on background conditions and geomorphic history. A change in bedload transport rating curves for channels with mixed-size loads is commonly associated with a change in armoring and organization

The evolution of sediment waves in heterogeneous rivers

471

of surface particles, but some depth of winnowing or ﬁll (i.e., a small change in storage) is required to adjust the size composition of surface particles. A change in sediment storage is commonly associated with a change in channel morphology or local gradient, but these also inﬂuence transport capacity. Adjustments to supply commonly occur at different rates in the same channel and to different degrees in different channels. The texture and organization of particles on the bed surface can respond quickly to changes in load, but more time is needed to transfer enough bed material to signiﬁcantly deform a channel. However, many of these adjustments are more available to some channels than they are to others. For example, a channel with strong banks and a wide GSD may readily respond to a change in load by a change in armoring, while one with weak banks and a narrow GSD may respond with a change in width–depth ratio, channel pattern, or sinuosity. Extremal hypotheses offer a theoretical approach to mutual adjustments of channel conditions to sediment load. Although some extremal hypotheses result in similar predictions, a theory that channels adjust to maximize ﬂow resistance in order to provide the least energy to transport the supplied load offers a physically intuitive approach to explore models of channel adjustment (Eaton et al., 2004). This theory can be used to predict combinations of dimensionless variables specifying the conﬁguration of a channel in equilibrium, but it does not specify the processes and rates leading to optimal system-scale ﬂow resistance and therefore has a limited ability to predict trajectories of change in an actual river. Nevertheless, the theory assumes that transport capacity is not a static quantity or relation, but a dynamic outcome of all aspects of channel condition responding to changing sediment supply. Rob Ferguson’s assertion that immediate changes in bedload rating curves may be no more than changes in sediment caliber brings into question, what is sediment supply in a channel dominated by bed material load? A bed material slug does not directly equate to a change in sediment supply, because sediment supply is necessarily a rate (volume supplied to a section of river per unit time) rather than merely a volume (say, the mass of the slug). Many slugs are ﬁner grained than ambient bed material and can directly reduce bedload caliber downstream; most have enough ﬁnes exposed in new deposits to be selectively transported and create a ﬁne pulse downstream. It is difﬁcult to disentangle the inﬂuence of sediment supply rate and sediment caliber in natural rivers, but ﬂume experiments (Dietrich et al., 1989; Lisle et al., 1993) and numerical models (Parker et al., this volume) indicate an inﬂuence of sediment supply on armoring and transport rates independent of sediment caliber. How then does one measure sediment supply? Sediment supply is commonly represented as bedload rating curves, and these show shifts in response to sediment inputs (see also, Sawada et al., 1983; D’Agostino and Lenzi, 1999). However, such direct measurements, as difﬁcult as they are, represent only one section of river. At any section, sediment supply is the product of mutual adjustments of topography, ﬂow, and sediment transport from the sequence of reaches upstream, each receiving sediment inputs and transferring some downstream. Therefore, to expand an evaluation of sediment supply from a river section to larger spatial and temporal scales in dynamic systems requires sediment routing schemes based on sound theory and good data. Ian Reid’s call for investigating hydraulic responses underlying shifts in rating curves would be vital in formulating such schemes.

472

T.E. Lisle

References D’Agostino, V., Lenzi, M.A., 1999. Bedload transport in the instrumented catchment of the Rio Cordon Part II: analysis of the bedload rate. Catena 36, 191–204. Dietrich, W.E., Kirchner, J.W., Ikeda, H., Iseya, F., 1989. Sediment supply and the development of the coarse surface layer in gravel-bedded rivers. Nature 340, 215–217. Eaton, B., Church, M., Millar, R.G., 2004. Rational regime model of alluvial channel morphology and response. Earth Surf. Process. Landf. 29, 511–530. Lisle, T.E., Iseya, F., Ikeda, H., 1993. Response of a channel with alternate bars to a decrease in supply of mixed-size bedload: a ﬂume experiment. Water Resour. Res. 29, 3623–3629. Parker et al., this volume. Adjustment of the bed surface size distribution of gravel-bed rivers in response to cycled hydrographs. Sawada, T., Ashida, k., Takahashi, T., 1983. Relationship between channel pattern and sediment transport in a steep gravel bed river. Z. Geomorph. N.F. 46, 55–66.

Gravel-Bed Rivers VI: From Process Understanding to River Restoration H. Habersack, H. Pie´gay, M. Rinaldi, Editors r 2008 Elsevier B.V. All rights reserved.

473

18 Sediment storage and transport in coarse bed streams: scale considerations Marwan A. Hassan, Bonnie J. Smith, Dan L. Hogan, David S. Luzi, Andre E. Zimmermann and Brett C. Eaton

Abstract Bedforms in gravel-bed rivers range in size from 102 to 103 m and have a wide range of affects on sediment transport and channel stability. To study the affect of these bedforms on sediment transport a classiﬁcation scheme is proposed that breaks the bedforms into micro, meso, macro and megaforms. Then using two creeks in British Columbia and one in California, the trend of sediment transport rates is related to the bed state, sediment and wood storage and the associated bedforms are discussed. At small spatial and temporal scales (micro/mesoform scales), variability in sediment transport rate can be ascribed to the changing state of the bed, which largely depends on the sediment supply regime. Stabilizing bedforms develop when sediment supply is low, and reduces the depth of the bed active layer and the mobility of the grains, thereby decreasing the sediment transport rate. Sediment rating relations in a low sediment supply channel are steep and are believed to be extremely sensitive to small changes in sediment supply and ﬂow regime when compared to higher sediment supply systems. High sediment supply suppresses the development of stabilizing bedforms and increases the mobility of grains and the depth of the active layer. At the reach scale (macroform scale), high sediment mobility is shown to produce complex cycles of aggradation and degradation that can persist for decades. Inchannel woody debris can strongly inﬂuence the timing and magnitude of these aggradation–degradation cycles and also has an important effect on the development of megaforms. 1.

Introduction

In the extensive literature on sediment transport in rivers, the predominant viewpoint is that the rate of sediment transport at any time is a function of hydraulic and sedimentological variables. In this framework, mass transport over a period of time is E-mail address: [email protected] (M.A. Hassan) ISSN: 0928-2025

DOI: 10.1016/S0928-2025(07)11137-8

474

M. A. Hassan et al.

calculated by integration over the period when ﬂows are competent to move the bed. The various functional relations presented for the calculation of sediment transport rates are based on dimensional analysis, stochastic or deterministic methods, and calibrated to some extent by laboratory and/or ﬁeld data. Predictions of sediment transport rate via hydraulically based functional relations are often more than an order of magnitude different than measured rates. These discrepancies have been explained by bed surface armouring and low sediment availability and the assumptions that underlie the models: uniform sediment, unconstrained movement of sediment and little consideration of the role of sediment supply, storage within the channel and sediment mobility. The effect of these interactions are presumably large, since transport rates are reported to vary by more than two orders of magnitude at constant ﬂow (e.g., Hayward, 1980; Jackson and Beschta, 1982). Coarse sediment, which comprises the bed material load, enters the stream channels episodically from adjacent slopes or upstream tributaries and often results in a step change in channel morphology and sediment transport patterns. Sediment is either initially deposited within the channel (Goff and Ashmore, 1994; Lane et al., 1995; Reid and Dunne, 2003) to be remobilized and moved onward by ﬂuvial processes at a later stage (e.g., Beschta, 1983; Sutherland et al., 2002), or it is immediately transported by ﬂuvial processes some distance downstream, where it is deposited, often behind obstacles to form sediment wedges and other sedimentation features. In all cases, the sediment inputs modify the channel morphology and bed surface composition (texture and structure), which affects the local sediment transport rate by altering sediment mobility and/or the distribution of shear stress acting on the bed. Evacuation of sediment stored in the channel following one of these episodic input events depends on ﬂood history (i.e., magnitude, duration and sequence), sediment characteristics and the sediment supply history. The result is that the temporal and spatial variation in the amount of within-channel sediment storage depends on the supply from external sources (e.g., Swanson et al., 1982b; Benda, 1990). Consequently, ﬂow events of the same magnitude and duration may produce different channel morphologies, bed surface texture and structure and sediment mobility (e.g., Bufﬁngton and Montgomery, 1999; Lisle et al., 2001). Streams with a relatively large sediment supply, and associated volume of in-channel sediment, typically have texturally ﬁner surfaces, poorly developed surface structures and higher sediment transport rates for a given discharge than channels with the same slope and lower sediment supplies (e.g., Lisle and Madej, 1992). When the sediment supply is low the development of a well-structured, coarse-textured bed signiﬁcantly reduces the sediment transport rates (Parker et al., 1982; Dietrich et al., 1989; Church et al., 1998; Hassan and Church, 2000; Ryan, 2001; Church and Hassan, 2002, Hassan and Woodsmith, 2004). Depending on the scale of the storage element and position within the stream network, sediment may be stored for periods ranging from less than a year to decades or even centuries (Dietrich et al., 1982; Swanson et al., 1982a; Kelsey et al., 1987; Madej, 1987, Madej and Ozaki, 1996). The transport capacity of a channel may appear constant at both short (o100 years) and very long timescales (4103 years), but it is clearly dynamic at intermediate timescales that correspond with the passage of sediment waves that cause ﬂuctuations in channel storage (Lisle and Church, 2002;

Sediment storage and transport in coarse bed streams

475

Lisle and Smith, 2003). At these intermediate timescales, Lisle and Church (2002) asserted that sediment transport capacity responds to changes in sediment supply and storage. For a given rate of sediment supply, the capacity regulates changes in sediment storage within a reach and the amount of sediment transferred to the next downstream segment. Lisle and Church (2002) showed an exponential decrease in the stored sediment with time implying that a linear relation may exist between transport capacity and storage. Further research has shown that a linear transport–storage function is only one possible curve and that more complex relations exist following severe aggradation that alters the channel pattern and that fundamentally different curves exist for relations between transport and storage during aggradation and degradation (Lisle and Smith, 2003; Smith, 2004). In the preceding discussion it has been emphasized that bed state and sediment storage play an important role in regulating sediment transport. What has not, however, been emphasized is the scale at which these factors are most evident. For instance, the concept of storage as a regulator of sediment supply applies at longer time scales and over larger spatial areas than have been investigated in many sediment transport studies. There is a need to illustrate with ﬁeld-based studies the time and spatial scales that are most appropriate when utilizing sediment storage functions. Furthermore, much of the research on sediment storage functions to date has been based on ﬂume studies and long-term, large-scale ﬁeld investigations are needed. Finally while bed structuring has been shown to inﬂuence sediment transport rates, its prevalence in different regimes (e.g., arid versus humid) and temporal stability have not been discriminately described. The primary objective of this paper is to examine temporal and spatial scales of within-channel sediment supply (availability), storage and mobility. We focus on intermediate sized streams (as deﬁned by Church, 1992), wherein inputs of woody debris and the development of grain-scale bedforms can both signiﬁcantly affect sediment transport patterns and channel morphology. In particular, we introduce a bedform classiﬁcation scheme that covers elements ranging in scale from 102 to 103 m. Three ﬁeld studies are then introduced and data from these studies are used to illustrate the effect of bed surface structuring on sediment transport rates as well as long-term sediment storage functions and the importance of large woody debris (LWD) as a regulator of sediment storage.

2.

Bedform classiﬁcation

While a number of authors have developed bedform classiﬁcation schematics, a classiﬁcation scheme that extends across the full range of scales over which bedforms are found in gravel-bed rivers does not exist. Such a scheme would help practitioners evaluate the space and time scales at which different bedforms are best studied and provide a template for discussion. Furthermore, the classiﬁcation scheme can improve our understanding of sediment exchange dynamics between different storage elements and thereby improve sediment routing models. The classiﬁcation schematic developed here is partly based on Lewin (1978), Church and Jones (1982) and Hassan (2005) and considers the channel bed of gravel-bed streams to be composed of

476

M. A. Hassan et al.

storage and resistant elements that can be classiﬁed as microforms, mesoforms, macroforms or megaforms (Fig. 18.1). At the channel unit scale (usually one to a few channel widths in length) microforms can occur depending on the bed state. Church (1978) deﬁnes bed state as either ‘‘overloose’’, ‘‘normally loose’’ or ‘‘underloose’’: overloose boundary includes open packed material and dilatant sediments (no/few microforms); a normally loose boundary consists of material resting in a nondispersed state without imbrication and with random packing (some microforms); and an underloose boundary is composed of close packed or a structured surface, including armoured surfaces and many microforms. The underloose state is typical of low sediment supply systems and overloose state is typical of high sediment supply regimes. A low sediment transport regime is often characterized by a coarse-grained and poorly sorted channel bed with imbricated structures. Large grains tend to be relatively over exposed while the smaller grains are hidden, as such, bed mobility is highly inﬂuenced by the relative exposure of particles within the bed. In addition to microforms, in low transport regimes, mesoforms composed of clusters, ribs and stone cells are also frequently well developed, thereby reducing the sediment mobility by increasing the threshold of entrainment by dissipating energy due to ﬂow resistance. Mesoforms are composed of two groups, at smaller scales ribs and stone cells scale with the grain size and at larger scales cascades, step-pool and rifﬂe-pool sequences scale with the channel width. Small-scale mesoforms (e.g., ribs, stone cells) evolve in response to changes in the ﬂow regime and sediment supply, and their existence indicates a quasi-stable bed with limited local in-channel sediment availability. For example, the size distribution of stone cells in Harris Creek, a cobble-bed stream, clearly varies from year to year (Fig. 18.2a), as does the proportion of the total bed area inﬂuenced by them (Fig. 18.2b). When the bed is well armoured, the surface is covered by a dense network of smaller cells (data from 1989 in Fig. 18.2a). Based on ﬂume experiments modelling Harris Creek, when the sediment supply is increased the surface exhibits stone cells that are larger but which stabilize a lower proportion of the total stream bed (Hassan and Church, 2000). Spatial variations in the proportion of area covered by surface structure are also characteristic of mesoforms (Fig. 18.2b). Within the same hydrological regime, streams with large sediment supply are likely to have a ﬁner surface and less-developed surface structures. When present, they are usually solitary features sparsely distributed over the bed surface, or are developed in association with some larger-scale bedform that alters the local sediment transport conditions. Similarly, due to their relatively large sediment supply and ﬂashy storm hydrographs, arid streams are likely to have less-developed armoured surface than their nival counterparts (Schick et al., 1987; Laronne et al., 1994; Reid and Laronne, 1995; Parker et al., 2003; Hassan et al., 2006; Fig. 18.3). This bed surface response is related to the local sediment transport rates and is expected to respond over short timescale (ﬂood timescale). Where wood is present, the channel may become starved downstream of the wood, promoting microform development, while at the same time becoming inundated with sediment upstream reducing or eliminating microforms. While microform features may persist for several years, the individual structures are thought to be transient.

Sediment storage and transport in coarse bed streams 477

Figure 18.1. Hierarchical bedform classiﬁcation. D is the average clast size of bedform, W the channel width, d the mean depth, te the event time. This classiﬁcation is partly based on Lewin (1978), Church and Jones (1982), and Hassan (2005). (Modiﬁed by permission of American Geophysical Union.)

M. A. Hassan et al.

478 60 (a)

Proportion

0.4

Structure area (%)

0.5

1989 1993

0.3 0.2 0.1 0.0

0.0

1.0 2.0 3.0 4.0 Structure length (m)

5.0

1989 Reach1

(b) 40

1993 Reach1 20 1993 Bar 0 0

5 10 15 20 Downstream distance (m)

25

30

1000

Low 6 4

D50s (mm)

Snow melt Humid Arid

2

100

1 High

Sediment supply

Figure 18.2. (a) Histograms of Harris Creek structure dimensions measured parallel to ﬂow direction. (b) Percentage of the area cover by structures in Harris Creek.

Snowmelt

10

1

Flash flood

1

10

100

D50sub (mm) Figure 18.3. Median size of surface and subsurface of ephemeral, snowmelt, humid and arid streams. (Data complied by Hassan et. al., 2006, Reproduced by permission of American Geophysical Union.)

At the spatial scale of channel reaches (usually 410 channel widths), large-scale mesoforms develop that span the stream channel. Large-scale mesoforms include step-pool units, cascades and rifﬂe-pool sequences and commonly persist for decades. The primary sediment storage mesoforms in intermediate streams (for deﬁnition, see Church, 1992) are bars and sediment wedges. Bars develop in association with rifﬂepool sequences or LWD pieces and sediment wedges deposited upstream of a downstream rise in base level due to factors such as a channel spanning log jams or a landslide entering the stream. The largest elements in Fig. 18.1 (macroforms and megaforms) function primarily as mid to long-term (decades-centuries) sediment stores. While, these features tend to last longer and cover a larger spatial extent in large streams, they are also present in small to intermediate streams, particularly those with a relatively large supply of sediment. Local inputs of LWD pieces affect the local sediment ﬂux. LWD that spans a signiﬁcant part of the channel often results in local aggradation upstream, leading to

Sediment storage and transport in coarse bed streams Degrading

D3

D2

Stable

D1

479 Aggrading

A1

A2

A3

(+) (+) (+) (+)

Step-Pool

D/d D/Wb s Confinement

Cascade-Pool

Riffle-Pool

Supply limited

Transport (-) (-) (-) (-) limited

Figure 18.4. Channel morphology matrix showing levels of disturbance. (Modiﬁed from Anonymous, 1996, with permission of American Geophysical Union.)

ﬁning and sediment starvation downstream, producing armouring (e.g., Hogan et al., 1998; Bufﬁngton and Montgomery, 1999). The inﬂuence LWD has on the within-channel sediment depends in large part on how high the sediment ﬂux is, and whether the channel reach is degrading, stable or aggrading (see Fig. 18.4). When the sediment supply rate is high, the channel aggrades, resulting in a morphology characterized by a series of sediment accumulation zones associated with channel-spanning LWD that has been partially buried (and thereby stabilized) by sediment. As the sediment supply is reduced or the LWD trap changes, sediment no longer accumulates behind LWD pieces. Instead, scour occurs beneath and around the ends of the LWD, exposing the LWD to greater hydraulic thrust and causing much of the wood to swing downstream, parallel to the ﬂow where it does not interact with the sediment transport ﬁeld (e.g., Hogan and Ward, 1997). Since sediment supply is highly episodic, channel morphology often cycles within or between the various morphologic states shown in Fig. 18.4. Aggradation–degradation cycles may persist and inﬂuence ﬁsh habitat for over half a century (Hogan and Ward, 1997; Hogan et al., 1998), and may comprise various changes to the bed, banks and/or function of LWD.

M. A. Hassan et al.

480

In general, all the elements described in Fig. 18.1 are present and functioning to varying degrees within a stream channel at any given time. Their spatial and temporal distribution within a channel changes signiﬁcantly as a function of local sediment storage and/or supply. Changes in sediment and LWD supply produces changes in the average bed surface texture and bar and wedge magnitude and frequency. Fig. 18.1 provides a classiﬁcation scheme that can be used to study sediment dynamics in ﬂuvial systems at different temporal and spatial scales. The data used to examine these concepts are derived from intermediate size, gravel-bed streams with a rifﬂe-pool morphology. Intermediate size streams were chosen because they contain all scales of sedimentation elements and required measurements have been collected over appropriate time and spatial scales.

3.

Field site description

Data used in this paper come from various studies reporting sediment transport measurements, detailed topographic surveys, geomorphic mapping and LWD inventories for streams having a wide range of bed texture, structure, channel morphology and ﬂow regime. Table 18.1 presents general characteristics of the three primary case studies Carnation Creek, Harris Creek and Tom McDonald. Carnation Creek drains a 11 km2 watershed on the west coast of Vancouver Island, British Columbia. The watershed is subjected to frequent rainstorms during the autumn and winter (Hartman and Scrivener, 1990). The bed material is mobile Table 18.1.

Summary characteristics of Carnation Creek, Harris Creek, and Tom McDonald.

Drainage area (km2) Bedrock Land use Annual precipitation (mm) Mean Annual ﬂood (m3/s) Unit mean annual ﬂood (m3/s/km2) Stream gradient (m/m) Channel width (m) Length of channel studied (m) Reach length relative to Bankfull width Armour ratio D50 subsurface (mm) Study period

Carnation Creek

Harris Creek

Tom McDonald

11 Volcanic Logged (since 1976) 3000 31 2.82 0.009 5–15 3100 310 1.5 25a 1971–present

187 Volcanic Logged (1950s)

18 Metamorphic Logged (1930–1950) 2000 3.6 0.20 0.006 10 10 1 1.3 16b 1985–1986

420 21 0.11 0.010 15–20 500 33 3.6 19a 1988–1995

Bankfull discharge for Tom McDonald Deﬁned as the ratio between D of the bed surface and D of the bed subsurface. 50 50 a

Based on bulk samples. Based on freeze core samples.

b

Sediment storage and transport in coarse bed streams

481

during ﬂows that occur several times a year. The surface has a low armour ratio and exhibits only solitary surface structures and irregular surface packing. Sediment and wood enters the channel through episodic debris ﬂows and bank erosion. Sediment is stored in bars and behind LWD jams. Low rates of bed load transport occur at discharge of about 3 m3/s and signiﬁcant transport rates, deﬁned as 1 kg/m/min, occur at discharges greater than or equal to 10 m3/s in the stream (Tassone, 1987). Harris Creek drains 187 km2 watershed in the southern interior of British Columbia. Although the drainage area is 17 times larger than Carnation Creek, its average bankfull width is similar due to the much drier interior climate. The stream ﬂow regime is dominated by snowmelt in May and June, and as a result, large ﬂood events outside the snowmelt period are not common; the mean annual ﬂood is about 19 m3/s. The bed is well armoured and large particles on the bed form reticulate networks (stone cells on Fig. 18.1) that increase hydraulic resistance to ﬂow and increase channel stability (Church et al., 1998). Most of the mobile sediment is delivered from upstream, where material is being released from eroding alluvial banks in laterally unstable reaches around broken LWD jams, and from sites where chronic debris raveling occurs from high banks in Pleistocene sediments (Ryder and Fletcher, 1991; Church and Hassan, 2005). Storage areas include bars, tributary alluvial fans and LWD jams. Tom McDonald Creek drains a 18 km2 watershed in north coastal California. The watershed is subjected to frequent rainstorms during the autumn and winter. Bankfull discharge at the study site is 3.6 m3/s. The largest peak ﬂow for which we have data is estimated at 25 m3/s, which is approximately twice the estimated regional mean annual ﬂood for drainage basins of similar size (Smith, 1990). The study reach consists of a single pool, including its head, centre and tail set within a rifﬂe-pool morphology. Gravel and cobbles dominate the pool head, sand covers most of the pool centre and pebbles dominate the pool tail. The length of the pool is approximately equal to the bankfull width of the channel (10 m). Sediment transport estimates at these sites are derived from observations using tracers, bedload sampling (Helley–Smith and Arnhem) or pit traps. Short- and longterm changes in sediment storage have been estimated from repeated channel surveys taken during ﬂoods, between ﬂoods and annually. Since the methods used for estimating sediment transport and storage have differing inherent biases, there is a degree of uncertainty in our analyses. However, we believe that the general trends in the data are realistic, even if the data itself may contain some biases.

4.

Bed state and sediment mobility

At small spatial and temporal scales, variability in sediment transport rate can be ascribed to the changing state of the bed and, in turn, the bed state largely depends on the local sediment supply regime (cf. Fig. 18.1). To demonstrate the inﬂuence of the bed state on sediment mobility, we compare two streams: Carnation Creek, where the dominant bed state is normally loose, and Harris Creek where the bed is dominantly underloose. One might suppose that since Harris Creek has a snowmelt dominated regime and Carnation Creek has a rain dominated regime that the

482

M. A. Hassan et al.

hydrology would signiﬁcantly inﬂuence the bed state. This must be true, to some extent, but since the contrast in sediment supply dynamics in the two systems appears to be greater than the hydrologic differences (they both transport sediment, on average, for the same duration per year) we attribute the observed differences in bed state to differences in sediment supply. Two examples (for more rating relations see Tassone, 1987 and Hassan and Church, 2001) of total load rating relations for Carnation Creek and Harris Creek (Fig. 18.5a) are used here to demonstrate the inﬂuence of the bed state on sediment transport. Hassan and Church (2001) demonstrate that the observed transport rates in Harris Creek are extremely low. The rating relations they present are very steep (the typical relation exponent (b in Fig. 18.5a) ranged between 4 and 14) and extremely sensitive to small changes in sediment supply and ﬂow regime. Gravel starts moving at about 4.5 m3/s but fractions larger than D80 are rarely mobile. Hassan and Church (2001) describe two phases of transport based on the data; sand transport over static bed (stage I) and partial transport of gravels at higher ﬂows (stage II). Full mobility of most sizes found on Harris Creek bed surface (called stage III) was not observed during 5 years of observations. Measured unit sediment transport rates in Carnation Creek are two to three orders of magnitude higher than those of Harris Creek at comparable discharge values and the rating relation is less sensitive and less steep, the relation exponent typically being less than 4 (Fig. 18.5a). Gravel in Carnation Creek starts moving at about 3 m3/s and the largest fractions in the bed are mobile annually. Bedload measurements show that both stage II and stage III were observed in Carnation Creek (Fig. 18.5a). To more systematically compare the difference in sediment mobility between the two creeks, we analyzed the fractional transport rate following Wilcock and McArdell (1993) and the proportion of tracers that moved relative to the total population of tracers in each size fraction (see Church and Hassan, 2002; Haschenburger and Wilcock, 2003). Fractional transport rates in Harris Creek are calculated from samples collected in pit traps and therefore a complete analysis of mobility of all grain sizes is possible. Transport rates in Carnation Creek were measured with an Arnhem bedload sampler which physically limits the size of the grains that can be sampled (o64 mm) and reduced sampling efﬁciency for grains larger than 32 mm. Transport rates in Harris Creek are up to three orders of magnitude below the reference transport rate suggested in the literature (W* ¼ 0.002) (e.g., Parker et al., 1982; Wilcock and McArdell, 1993) (Fig. 18.5b). All fractional transport curves, scaled by the subsurface grain size distribution, show a break in the slope where transport rate begins to decline with increasing grain size, which corresponds to a shift from full to partial mobility (Fig. 18.5c). At low ﬂows no size fraction in Harris Creek is fully mobile while at the intermediate ﬂows (47 m3/s) sand is fully mobile and larger material remains partially mobile. At high ﬂows (412 m3/s) the division between full and partially mobile size fractions occurs around 16 mm, which is ﬁner than the median size of the subsurface material (D50 of the subsurface is 19 mm). Even at ﬂows as high as the mean annual ﬂood, scour is sporadic and limited in depth: as a result, the framework cobble and gravel remains in place and the largest material scarcely moves. These ﬁndings are supported by results of the tracer study during ﬂood events in 1989, 1990 and 1991 (Fig. 18.5e). On an average, 60% of the

10,000

1000

(a)

(b)

Carnation Creek

Stage III Stage II

Carnation Creek b = 2.74

10

1

Stage II

0.1

Harris Creek b = 13.18

0.01

Stage I

100

W*

1

0.01

Harris Creek

0.0001 2

10

10

50

3

100

Shear stress (Pa) 10,000

100

(d) Carnation Creek

D50 Surface

(c) Harris Creek 10.0

Fully mobile

1000

Fully mobile Partially mobile

0.100

100

0.010

(m /s) Q=14.75 Q=12.62 Q=9.18 Q=5.00 Q=3.81

0.001 0.1

1.0

Partially mobile

3

10.0

10

3

(m /s) Q=15.3 Q=14.2 Q=12.9

1

100.0

1.0

Particle size (mm) (e) Harris Creek

100

(f) Carnation Creek

100.0

(g) Harris Creek

0.6

(h) Carnation Creek

19

Proportion

91 19

D50 Subsurface

D50 Surface

9

D50 Surface

20

198

D50 Subsurface

40

89

60 1990 1991

% moved

10.0

Particle size (mm)

80

0

Q=8.4 Q=6.5

D50 Surface

1.00

D50 Subsurface

Scaled fractional transport rate (g/m s)

Discharge (m /s)

D50 Subsurface

0.001

Scaled fractional transport rate (g/m s)

Transport rate (g/m s)

100

16

64

0.3

Exposed

0.0

4

16

64

256 1024 4

Particle size (mm)

256

1024

0

3

6

9

0

3

6

9

Layer number

Figure 18.5. (a) Bedload rating curves derived from trap observation in Harris Creek and Arnhem sampler in Carnation Creek. (b) Scaled fractional transport rate versus shear stress for Harris Creek (after Church and Hassan, 2002) and Carnation Creek. Scaled by the subsurface grain size distribution (c and d) Scaled fraction transport rate in Harris Creek versus particle size (after Church and Hassan, 2002) and Carnation Creek, respectively. (e and f) Percent of moved tracers as a function of grain size in Harris Creek (after Church and Hassan, 2002) and in Carnation Creek, respectively. (g and h) Burial depth (layer number) distribution after individual ﬂow events in Harris Creek and Carnation Creek, respectively (after Hassan and Church, 1994). Cross-hatching denotes exposed particles. (Reproduced by permission of American Geophysical Union.)

484

M. A. Hassan et al.

entire tracer population moved during the 1989 freshet, 88% during 1990 and only 36% in 1991. A plot of fractional mobility shows a sharp decline in mobility with increasing particle size for all three seasons. About 70% of sizes ﬁner than the median size of the bed material moved during both 1989 and 1991 ﬂood seasons. In general, measurements from both the pit trap and tracer studies show full mobility was approached only for sizes of less than about 16 mm, while the vast majority of the bed surface grain sizes were only partially mobile. Fractional transport rates in Carnation Creek tend to be one to three orders of magnitude higher than the reference transport rate (Fig. 18.5b). The break in the slope for each curve in Fig. 18.5d represents the upper limit of full mobility for the corresponding discharge. At low ﬂows, full mobility is limited to grain sizes less than about 2 mm, while grains up to 8 mm are fully mobile at intermediate discharges. For high ﬂows, the largest grain size class (16–32 mm) that can be efﬁciently sampled with the Arnhem sampler is fully mobile. Results from the tracer study in Carnation Creek for ﬂood events in 1989 and 1991 show that almost all particles ﬁner than the median size of the bed material moved during 1989 and 1991 (Fig. 18.5f). Additionally, all the tracers ﬁner than the median size of the surface material moved during 1991. For material coarser than the median size of the bed surface material, the two seasons show a decline in the mobility rates with the increasing particle size but at much lower rate than that obtained for Harris Creek. In 1989, the full mobility for Carnation Creek approached the median size of the bed material while in 1991 sizes larger than the median size of the surface material were fully mobile. Furthermore, a signiﬁcant proportion (450%) of the largest bed material was mobile during both seasons. To better understand sediment transport dynamics in relation to bed state, we explore burial depth as an indicator of relative bed material mobility between Harris Creek and Carnation Creek. In Fig. 18.5g it is seen that a large proportion (450%) of the tracers in Harris Creek remained on the surface during the largest ﬂood event in the 3-year record (i.e., 1990); only small proportions were deeply buried. The rapid decline in proportion of tracers with increasing depth indicates that the scour during the ﬂood event was shallow and localized, and that the surface framework of particles remained in place during the ﬂood event. In contrast, less than 10% of the tracers in Carnation Creek remained on the bed surface while some were deeply buried (Fig. 18.5h). The relatively gentle decay of burial depth indicates that the bed surface was substantially disrupted during the ﬂood event. The difference in burial depth, transport rates and the extent of full/partial mobility cannot be explained simply by differences in the shear stress in the two streams. The dimensionless shear stress (based on the subsurface material) in Harris Creek was 0.071, and in Carnation Creek it was 0.070. However, when we substitute the surface median grain size, the dimensionless shear stress in Harris Creek falls to 0.044 and in Carnation Creek to only 0.059. Furthermore the surface structures that are pervasive on the bed of Harris Creek probably further delayed the entrainment of the surface material. The differences in particle mobility and, as a direct consequence, in sediment transport rate were, we believe, strongly conditioned by the assemblage of microforms and mesoforms that are present in Harris Creek and not in Carnation Creek. We believe that these surface structures developed in response to a relatively

Sediment storage and transport in coarse bed streams

485

low sediment supply rate as supported by the experiments of Hassan and Church (2000), while the high sediment supply rate inhibited the development of similar structures in Carnation Creek. Ultimately a difference in sediment supply from hillslopes must be responsible, as over time, if supply conditions were identical Carnation Creek would be expected to structure to a state similar to that of Harris Creek (as predicted by Fig. 18.4).

5.

Storage and mobility at the channel unit scale

At the macroform scale, investigations of sediment transport can be completed at the channel unit or reach scale, and examples of both are given. At the channel unit scale temporal patterns of sediment storage-transport during a ﬂood event are demonstrated using data collected from Tom McDonald Creek (Hassan and Woodsmith, 2004). During the study period, four small events (obankfull discharge) and one major event (7 times bankfull discharge, 25 m3/s) were recorded (Fig. 18.6a). Topographic surveys and bed load sampling (using a Helley–Smith sampler) were conducted along three transects (pool head, pool centre and pool tail) repeatedly during each ﬂood event. Net changes in sediment storage volume at each transect are typically plotted against time (Fig. 18.6b). Information from Fig. 18.6b can be used to produce a transport–storage relation (Lisle and Church, 2002) where time is implied (Fig. 18.6c). The advantage of presenting the data in the latter format is that magnitudes and cycles of aggradation and degradation can be easily identiﬁed. Data in the plot represents a measurement of initial storage volume and subsequent transport within a 1-day interval. Arrows (Fig. 18.6c) indicate the direction of increasing time. Each transect illustrates spatial variability of storage and net accumulation of bed material (Fig. 18.6b). Prior to the largest ﬂood event (day 775), all transects within the pool show minimal changes (Fig. 18.6c). During the rising limb of the large ﬂood, net ﬁll was measured at all transects with relatively small amounts of ﬁll at the pool centre and tail and signiﬁcant ﬁll at the pool head. During the falling limb and subsequent ﬂows, the behaviour of the transects diverge. Net change at the pool tail was minimal while the pool head ﬁlled and the pool centre degraded at constant rates. The bed material transport–storage relation based on data from all three transects (Fig. 18.6c) shows a small net increase in storage volume through the peak of the large ﬂood. Storage volumes then increased rapidly over a 2-day period during the falling limb of the large ﬂood when the ﬂow was well above bankfull discharge. Minor degradation was observed during the fourth small event. Transport–storage relations show small ﬂuctuations, on the order of the resolution of the survey, without a clear cyclic pattern prior to the large ﬂood. The large ﬂood initiated a large scale aggradation–degradation cycle that was not completed during the study period (1 year). Rating curves based on Helley–Smith measurements do not show distinct trends between the head, centre and tail (Figs.18.6d, e, f) and they do not reﬂect changes in local storage measured at each transects or the pool average. The sediment yield

M. A. Hassan et al.

486 (a) Storage volume (m3)

Discharge (m3/s)

30

20

10

0 700

750

800

850

Day

Before day 777 After day 777

(b)

40

Head

20

0

Tail Centre with bank collapse

-20 200

400

600

800

1000

Day

(c) Scour

Scaled inital storage volume 0

-20

Fill

Output rate (m3/day)

20

-40 -100

0

50

100

Storage volume (m3) 12.8

100.1 0.55

(d) Head

4.6

37.2

1.2

(e) Centre

9.2

68.5

(f) Tail

100

1 Before day 777 After day 777

QBF

Transport rate (g/m s)

Shear stress (Pa) 1.6 10,000

Before day 777 After day 777

Before day 777 After day 777

0.01 1

10

100 1

10

100 1

10

100

Discharge (m3/s)

Figure 18.6. Tom McDonald Creek – (a) Flow hydrograph at the study site for the 1986 season. (After Hassan and Woodsmith, 2004.) (b) Temporal variation in the volume of sediment storage during 1985–1986 seasons. (c) Temporal variation in stored sediment versus output rate. (d, e, and f ) Bedload rating curves derived from Helley–Smith sampler. The 777 day indicates major change in bed elevation. (Reproduced with permission from Elsevier, 2004.)

estimated from Helley–Smith measurements is an order of magnitude larger (head ¼ 5686 m3; centre ¼ 7368 m3; tail ¼ 4248 m3) relative to the change in storage measured during the same period (79 m3). This result is likely due to the sediment transport step length being much longer than the study area, thus much of the sediment being captured in the Helly–Smith samplers did not originate from within the study area. This study clearly illustrates the disconnect that may occur if the scale of investigation, is smaller than the scale of the phenomenon being investigated.

Sediment storage and transport in coarse bed streams 6.

487

Sediment storage at the reach scale

A larger spatial and temporal scale is more appropriate for studying macroform bed elements. Data from Carnation Creek illustrates the temporal variability and trends of sediment transport over a 34-year period. Since 1971 the British Columbia Ministry of Forests has maintained eight study areas (a total of approximately 500 m) along the main channel of Carnation Creek, each of which has between 18 and 36 cross-sections. Using the cross-sections, spaced on average 3 m apart, annual measurements of stream bed topography were completed to track changes in the channel morphology. These repeat cross sections are used to estimate long-term sediment ﬂux occurring through each study area. Volumes were computed by constructing Triangular Irregular Networks (TIN) from channel cross-section survey data from each year. The TIN was converted to a 5 cm DEM and the DEM from the year of interest was subtracted from the 1971 DEM (the ﬁrst year of record). The calculated change in storage since 1971 was then compared to the change in storage calculated for the previous year to determine how much sediment was lost or gained during the winter. Errors in the annual survey are generally less than about 5 cm or 50 m3 per study site. Reach scale ﬂuxes appear to be cyclic, involving a consistent trend of channel aggradation (or degradation) followed by a degradational (or aggradational) trend. These trends are assumed to be reﬂective of the role of sediment supply and LWD in modulating the storage and release of sediment. Aggradation and degradation was not found across all study sites during the same years, nor were there any observable ﬂuxes of sediment from reach to reach. This suggests that aggradation–degradation cycles at each reach have local controls in additions to upstream sediment supply. Storage cycles have a range of characteristic magnitudes, which we classify as small (100–400 m3) and large (4400 m3). Data from two of the study reaches are used to illustrate the spatial and temporal changes in sediment ﬂux and storage over several decades. These cycles are hypothesized to be controlled by a combination of sediment supply, channel morphology and LWD. Large–scale aggradation–degradation cycles are only found in reaches with channel spanning LWD jams which trap sediment and create reach scale sediment wedges. Small-scale aggradation–degradation cycles are present within the record in all reaches and are hypothesized to correspond to the growth and lateral shifting of bars driven by changes in sediment supply and sedimentation associated with smaller wood accumulations. Study area VIII (Fig. 18.7) is the upstream-most zone considered here. It is situated downstream of a steep canyon (Fig. 18.7), and demonstrates the role of LWD in affecting spatiotemporal patterns of sediment storage. Study area VI is approximately 475 m downstream of study area VIII. While LWD is found in the channel as isolated wood pieces at study area VI, no channel spanning LWD jams have been documented within the reach in the last 32 years (Fig. 18.8). Both scales of aggradation–degradation cycles are evident in the data from reach VIII (Fig. 18.7a). The ﬁrst aggradation–degradation cycle (1971–1981) is small in size (270 m3) and roughly corresponds with the pre-jam phase and the early stages of jam development (jam adolescence, Luzi, 2000), as shown in Fig. 18.7b. In 1982, the LWD jam in Fig. 18.7b trapped a signiﬁcant amount of additional debris

488

M. A. Hassan et al.

(Fig. 18.7c): this additional accumulation deﬂected ﬂows against the right bank and caused substantial erosion of the bank and a net decline in sediment storage in the reach. Degradation stopped abruptly in 1984 when material that originated from a debris ﬂow in the canyon, displaced the existing LWD jam downstream (Powell, 1987), forming a channel wide barrier to downstream sediment transport (Fig. 18.7d) which initiated a large-scale aggradation–degradation cycle (1982–1991) (Fig. 18.7a). Sediment accumulated behind the LWD jam from 1983 to 1989 in the form of a large sediment wedge (Luzi, 2000). In 1990, the channel cut partially around the LWD jam and began to incise into the sediment wedge deposit. The majority of sediment removal occurred in 3 years (1989–1991), marking the end of the large–scale aggradation–degradation cycle. The volume of stored sediment in 1991 was 290 m3 greater than the pre-disturbance storage volume measured in 1971. A third small-scale aggradation–degradation cycle (100 m3), equivalent in size to the 1971–1982 cycles, occurred between 1991 and 1995. By 1995, the channel had incised vertically such that the jam only interacted with the channel during high ﬂows (Luzi, 2000). The channel degraded from 1995 to 2000 an additional 370 m3 and had a ﬁnal storage volume 36 m3 less than that measured in 1971. Four small-scale cycles of aggradation and degradation are identiﬁed in study area VI (Fig. 18.8a). The ﬁrst two cycles are similar in magnitude to those observed in study area VIII, spanning 5 and 12 years respectively (1971–1976 and 1976–1988). Cycle 3 is slightly larger than the earlier cycles and spans 15 years, and cycle 4 is nested within cycle 3 (Fig. 18.8a). Aggradation–degradation cycles 2 and 4 are separated by rapid aggradation from 1989–1991, and may correspond to the release of sediment from the LWD jam in study area VIII. However, there is not a similar signal at study area VII, located between study area VIII and VI. In 2003, there was about 50 m3 of sediment left in storage compared to 1971, which is within the errors associated with the surveys. These two study areas illustrate that the trajectory of change in transport and storage persist through multiple years at the reach scale. The small-scale cycles are the most common, are present in both study areas and are associated with sediment accumulation and release. Large-scale aggradation–degradation cycles are clearly controlled by the formation of LWD jams which trap sediment and create reach scale sediment wedges. LWD jams may also affect downstream reaches by interrupting small cycles by changing the sediment supply regime. Carnation Creek reaches have a level of disturbance between D1 and A3 on the conceptual matrix shown in Fig. 18.4 based on channel morphology that include mid-channel bars, LWD jams and poorly developed structures. If sediment supply is reduced in Carnation Creek, such as downstream of an intact recently formed LWD jam, the channel will degrade and develop surface armouring and structures (as shown by Haschenburger and Rice, 2004). This change in bed state will reduce sediment mobility and create small-scale cycles of aggradation–degradation. Under reduced sediment supply, Carnation Creek will likely shift to a level of disturbance between D2 and D3 (Fig. 18.4), similar to Harris Creek. A signiﬁcant increase in the sediment supply into Harris Creek will result in aggradation, development of more substantial bars, increased bank erosion and destruction of bed surface structure and

Sediment storage and transport in coarse bed streams

489

1000 (a)

800

1991-1995 1995-2003

200

1971 2003

0 -200

Degrade

400 4

3

2 1982-1991

1971-1982 1

Aggrade

Output rate (m3/yr)

600

-400 -600 -800 -400

-200

0

200

400

600

800

1000

1200

1400

1600

Storage volume (m3)

R-50

(b) 1980

R-50B

R-50A

(c) 1983

R-50

R-50B

0

metres 10

R-50

(d) 1986

(e) 2003

R-50A

Water surface LWD Top of bank Bottom of bank Bar

Figure 18.7. Carnation Creek – (a) Temporal variations in stored sediment versus output rate in study area VIII. (b, c, d, and e) Examples of temporal changes in LWD jam formation and channel morphology of study area VIII. Top and bottom of banks are indicated.

armouring. Likewise an increase in wood supply will lead to more log jams and intern more bars and sediment wedges. Under the increased sediment supply scenario, the sediment mobility will increase the magnitude of the aggradation– degradation cycles.

M. A. Hassan et al.

490

(a)

19711976

1971

Degrade

100

1998-2003

50

1

0

No change

3

-50

4

2003

-100

Aggrade

Output rate (m3/yr)

150

1991-1998

2

-150 19761988

-200

1989-1991

-250 -300 -150

-100

-50

0

50

100

150

200

250

300

350

Storage volume (m3)

(b) 1983

R-41b

R-41a

R-40b R-40c R-40a

(c) 2003

0

metres 10

Figure 18.8. (a) Temporal variation in stored sediment versus output in study area VI. (b and c) Examples of temporal changes in channel morphology of study area VI. Top and bottom of banks are indicated.

Sediment storage and transport in coarse bed streams 7.

491

Sediment storage at the channel scale: the role of LWD

In the previous sections we considered temporal and spatial patterns of sediment storage and transport at the channel unit and reach scales. The impact of LWD on channel characteristics has been demonstrated in study areas VI and VIII in Carnation Creek (Figs. 18.7 and 18.8). However, LWD jams are not isolated occurrences along the longitudinal proﬁle of the stream. The spatial and temporal pattern of sediment storage in association with both bars and LWD jams was documented along the entire mainstem of Carnation Creek. Longitudinal thalweg proﬁles, LWD jam age, volume, span, height and channel location, log steps, and individual LWD piece, length and mean diameter were inventoried in 1991 and 1999. The 3.2 km survey included 55 jams in 1991, and 48 jams in 1999. The impact of a LWD jam on channel morphology depends on its size relative to the bankfull channel dimensions as well as the level of jam development. Detailed longitudinal proﬁles from a selected section of channel are shown from the 1991 and 1999 surveys (Figs. 18.9a, b). Areas where bar and bank top elevation merge indicate an exhausted local channel storage capacity due to severe aggradation and are probable sites for overbank ﬂow during ﬂood events and potential channel migration/avulsion areas (Luzi, 2000). Large convexities exhibited by the longitudinal proﬁle are the result of sediment accumulation upstream and scour downstream of a LWD jam, these features are more pronounced around newly formed channels spanning LWD jams. The difference in thalweg elevation upstream and downstream of LWD jams becomes less pronounced with time as the trapping efﬁciency of jams decline (Fig. 18.9a). The total average LWD volume per bankfull interval was 19.7 m3 in 1991 and 22.5 m3 in 1999, with large volumes (about 70 m3) associated with jams (Figs. 18.9c, d). The percentage of LWD volume associated with jams was 90% during both 1991 and 1999. Although the LWD volume remains relatively constant, jam function (trapping sediment) changes over time. Jams directly inﬂuence the amount of sediment stored in the channel (Figs. 18.9e, f). Sediment volume was greatest at sites upstream of channel spanning jams. However, the temporal and spatial extent of the sediment accumulation is strongly dependent on jam age. Variation in the volume of sediment storage per bankfull interval in the 1991 survey is closely linked to its spatial proximity to jams (Fig. 18.9e). Twenty-one jams with associated upstream sediment wedges were identiﬁed in 1991. In total, these wedges accounted for 10,100 m3 of sediment, approximately 47% of the total sediment stored in bars (21,400 m3) based on bar depth estimates in conjunction with channel width measurements from the long proﬁle surveys. In 1999, 20 jams with upstream sediment wedges were identiﬁed, and total sediment storage was estimated at 7000 m3, accounting for 38% of total bar-stored sediment (18,400 m3). Nineteen of the jams associated with sediment wedges were the same for both survey years. At the channel scale in Carnation Creek, bars store more sediment than LWD jams. However, LWD jams store large volumes of sediment at a few, infrequent, locations. Conversely, bars store sediment in many small structures distributed somewhat evenly along the channel length (approximately 70 bars as opposed to far fewer than 20 intact jam locations).

M. A. Hassan et al.

492 J - 7yrs SA 8 J - 7yrs

(a) 1991

28

J - 7yrs

J - 7yrs

J - 9 yrs SA 6

24 0y J rs -1 5y rs

J - 15yrs

J

-1

J - 13yrs

bank top bar thalweg

Elevation (m)

20

LWD jam 16

28

SA 8 J - 15yrs J - 15yrs J - 15yrs J - 15yrs

(b) 1999

J - 17yrs

24

20

3y -2

J - 21yrs

J

J - 18yrs

rs

SA 6

16 2000

2200

2400

2600

2800

3000

Wood volume (m 3)

300

(c) 1991 200 100 0 600

Sediment storage volume (m3 )

(d) 1999

(e) 1991

(f) 1999

Bars LWD jam

400

200

0 2000

2400

2800

2000

2400

2800

Distance along thalweg (m) Figure 18.9. Carnation Creek – (a and b) Detailed longitudinal proﬁle of selected section of the 1991 and 1999 survey, respectively. (c and d) Within-channel LWD volume in 1991 and 1999. (e and f) Volume stored sediment behind LWD and bars (after Luzi, 2000). Bars show total storage, shading shows proportion associated with wood (grey) and bars (white). (Reproduced by permission of American Geophysical Union.)

Sediment storage and transport in coarse bed streams

493

Recently formed, intact jams can be very large. Hogan and Bird (1998) classify jams according to 11 attributes (e.g., height, width, sediment trapping efﬁciency, among others) and four size classes (micro, meso, macro and mega). They documented mega jams that inﬂuence channel morphology for over 100 bankfull width in length. This includes aggraded zones upstream of the barrier (upstream extent is a function of pre-jam formation channel gradient) but the most extensive changes occur downstream as a result of scour with no recruitment of sediment from upstream (unable to pass the intact jam). In Carnation Creek, the net loss of sediment supply from behind jams was approximately 5800 m3 between 1991 and 1999. Of this, an additional 3100 m3 was deposited at sites either immediately downstream of the jam or within jam complexes further downstream. It appears that smaller secondary jams are developed downstream as the mega jams begin to deteriorate; mega jams release debris that form jams (range in size from macro to micro) that then inﬂuences sediment storage at the reach and unit channel scales.

8.

Conclusions

To study the affect of bedforms on sediment transport in gravel-bed rivers a classiﬁcation scheme was proposed that breaks the bedforms into micro, meso, macro and megaforms. It was shown that at small spatial and temporal scales (micro/ mesoform scales), bed structures regulate within channel sediment supply and are strongly inﬂuenced by upstream sediment supply. Stabilizing bedforms develop when sediment supply is low, this reduces the depth of the bed active layer and the mobility of the grains, thereby decreasing the sediment transport rate. Sediment rating relations in a low sediment supply channel are shown to be steep and are believed to be extremely sensitive to small changes in sediment supply and ﬂow regime when compared to higher sediment supply systems. High sediment supply suppresses the development of stabilizing bedforms and increases the mobility of grains and the depth of the active layer. At the reach scale (macroform scale), using Carnation Creek data, it was shown that the supply of sediment and wood from upstream can inﬂuence the development of sediment stores, such as channel bars and sediment wedges, which in turn have a pronounced effect on downstream sediment transport rates. The accumulation and release of sediment from bars and sediment wedges created characteristic aggradation–degradation cycles that persist for multiple years and are believed to depend more on external sediment supply rate, bedform dynamics and the supply of LWD than on the hydraulic forcing. These results have obvious implications on the stability of rating curves which are based solely on stream ﬂow and neglect changes in sediment or wood supply. Rating curves may be stable for a constant sediment supply that produces small scale aggradation–degradation cycles, but may have signiﬁcant shifts for large-scale aggradation–degradation cycles. It is possible that the cyclic nature of aggradation–degradation may appear as actual shifts in the rating, or as scatter within the rating curve.

494

M. A. Hassan et al.

More careful attention to the full range of bedforms occurring across all scales will enable us to better predict sediment transport rates, particularly for intermediate to long time scales when storage and bed state conditions are unlikely to be stable. Incorporating such instabilities into multi-year sediment transport models remains an important challenge today.

Acknowledgements The research was funded by the Natural Sciences and Engineering Research Council of Canada through Research Grants to M. Church and M. Hassan and a graduate scholarship for Andre´ Zimmermann. This paper beneﬁted from many discussions with M. Church. His support and supply of resources has been unlimited over the years. Tom Lisle kindly reviewed a draft and provided comments that improved the paper. The data from Carnation Creek would not exist if it was not for the many, many people who have surveyed the channel for the last 30 years; we thank everyone who has put time and energy into data collection at Carnation Creek. In addition, Rick Woodsmith collected and supplied the data from Tom MacDonald Creek and many, many students of UBC collected and sieved the data from Harris Creek.

References Anonymous, 1996. Channel Assessment Procedure: Field Guidebook. Ministry of Forest and British Columbia Environment, Victoria, BC, 95pp. Benda, L., 1990. The Inﬂuence of debris ﬂows on channels and valley ﬂoors of the Oregon Coast Range, USA. Earth Surf. Process. Landf. 15, 457–466. Beschta, R.L., 1983. Channel changes following storm-induced hillslope erosion in the Upper Kowai Basin, Torlesse Range, New Zealand. J. Hydrol. (NZ) 22, 93–111. Bufﬁngton, J.M., Montgomery, D.R., 1999. Effects of supply on surface textures of gravel-bed rivers. Water Resour. Res. 35, 3523–3530. Church, M., 1978. Palaeohydraulic reconstructions from a Holocene valley ﬁll. In: Miall, A.D. (Ed.), Fluvial Sedimentology. Canadian Society of Petroleum Geologists, Mem. 5, Calgary, Canada, pp. 743–772. Church, M., 1992. Channel morphology and topology. In: Calow, C., Petts, G. (Eds), The Rivers Handbook. Blackwell, Oxford, Vol. 2, pp. 126–143. Church, M., Hassan, M.A., 2002. Mobility of bed material in Harris Creek. Water Resour. Res. 38, doi:10.1029/2001WR000753: 19-1–19-12. Church, M., Hassan, M.A., 2005. Upland gravel bed rivers with low sediment transport. In: Batalla, R., Garcia, C. (Eds), Catchment Dynamics and River Processes. Amsterdam, Elsevier. 141–168. Church, M., Hassan, M.A., Wolcott, J.F., 1998. Stabilizing self-organized structures in gravel-bed stream channels: Field and experimental observations. Water Resour. Res. 34, 3169–3179. Church, M., Jones, D., 1982. Channel bars in gravel bed rivers. Gravel-bed rivers. In: Hey, R.D., Bathurst, J.C., and Thorne, C.R. (Eds), Gravel Bed Rivers. Wiley, Chichester, U.K., pp. 291–338. Dietrich, W.E., Kirchner, J.W., Ikeda, H., Iseya, F., 1989. Sediment supply and the development of the coarse surface layer in gravel-bedded rivers. Nature 340, 215–217. Dietrich, W.W., Dunne, T., Humphrey, N.F., Reid, L.M., 1982. Construction of sediment budgets for drainage basins. In: Swanson, F.J., Janda, R.J., Dunne, T., Swanston, D.N. (Eds), Sediment Budgets and Routing in Forested Drainage Basins. United States Department of Agriculture, General Technical Report Number PNW-141, pp. 5–23.

Sediment storage and transport in coarse bed streams

495

Goff, J.R., Ashmore, P.E., 1994. Gravel transport and morphological change in braided Sunwapta river, Alberta, Canada. Earth Surf. Process. Landf. 19, 195–212. Hartman, G.F., Scrivener, J.C., 1990. Impacts of forestry practices on a coastal stream ecosystem, Carnation Creek, British Columbia. Can. Bull. Fish. Aquat. Sci. 223, 148. Haschenburger, J.K., Rice, S.P., 2004. Changes in woody debris and bed material texture in a gravel-bed channel. Geomorphology 60, 241–267. Haschenburger, J.K., Wilcock, P.R., 2003. Partial transport in a natural gravel bed channel. Water Resour. Res. 39, 1020, doi:10.1029/2002WR001532. Hassan, M.A., 2005. Characteristics of gravel bars in ephemeral streams. J. Sediment. Res. 75, 29–42. Hassan, M.A., Church, M., 2000. Experiments on surface structure and partial sediment transport on a gravel bed. Water Resour. Res. 36, 1885–1895. Hassan, M.A., Church, M., 2001. Rating bedload transport in Harris Creek: Seasonal and spatial variation over a cobble-gravel bed. Water Resour. Res. 37, 813–825. Hassan, M.A., Egozi, R., and Parker, G., 2006. Effect of hydrograph characteristics on vertical sorting in gravel-bed rivers: Humid versus arid environments. Water Resour. Res. 42, W09408, doi:10.1029/ 2005WR004707. Hassan, M.A., Woodsmith, R., 2004. Bedload transport in an obstructed-formed pool in a forested gravelbed stream. Geomorphology 58, 203–221. Hayward, J.A., 1980. Hydrology and stream sediments in a mountain catchment. Special Publication 17, Tussock Grasslands and Mountain Lands Institutes, Lincoln College, Canterbury, New Zealnad, 236pp. Hogan, D.L., Bird, S.A., 1998. Classiﬁcation and assessment of small coastal stream channels. In: Hogan, D.L., Tschaplinski, P.J. and Chatwin, S. (Eds), Carnation Creek and Queen Charlotte Islands, Fish/ Forestry Workshop: Applying 20 Years of Coastal Research to Management Solutions B.C. Ministry of Forests, Working Paper 41, pp. 189–200. Hogan, D.L., Bird, S.A., Hassan, M.A., 1998. Spatial and temporal evolution of small coastal gravel-bed streams: Inﬂuence of forest management on channel morphology and ﬁsh habitat. In: Klingeman, P.C., Beschta, R.L., Komar, P.D., and Bradley, J.B. (Eds), Gravel-bed Rivers in the Environment. Water Resources Publication Highlands Ranch, Colorado, pp. 365–392. Hogan, D.L., Ward, B.R., 1997. Watershed geomorphology and ﬁsh habitat. In: Slaney, P.A. and Zaldokas, D. (Eds), Fish Habitat Rehabilitation Procedures, Watershed Restoration Technical Circular No. 9, British Columbia Ministry of Environment, Land and Parks, Vancouver, British Columbia, pp. 2-1–2-18. Jackson, W.L., Beschta, R.L., 1982. A model of two-phase bedload transport in an Oregon Coast Range stream. Earth Surf. Process. Landf. 7, 517–527. Kelsey, H.M., Lamberson, R., Madej, M.A., 1987. Stochastic model for long-term transpor of stored sediment in a river channel. Water Resour. Res. 23, 1738–1750. Lane, S.N., Richards, K.S., Chandler, J.H., 1995. Morphological estimation of the time integrated bedload transport rate. Water Resour. Res. 31, 761–772. Laronne, J.B., Reid, I., Yitshak, Y., Frostick, L.E., 1994. The non-layering of gravel streambeds under ephemeral ﬂood regimes. J. Hydrol. 159, 353–363. Lewin, J., 1978. Floodplain geomorphology. Prog. Phys. Geogr. 2, 408–437. Lisle, T.E., Church, M., 2002. Sediment transport–storage for degrading gravel-bed channels. Water Resour. Res. 38, 1219. doi:1210.1029/2001WR001086. Lisle, T.E., Cui, Y., Parker, G., et al., 2001. The dominance of dispersion in the evolution of bed material waves in gravel bed rivers. Earth Surf. Process. Landf. 26, 1409–1420. Lisle, T.E., Madej, M.A., 1992. Spatial variation in armouring in a channel with high sediment supply. In: Billi, P., Hey, R.D., Thorne, C.R., and Tacconi, P. (Eds), Dynamics of Gravel-bed Rivers. Wiley and Sons, Chichester, UK, pp. 277–311. Lisle, T.E., Smith, B., 2003. Dynamic transport capacity in gravel-bed river systems. In: Proceeding of the International Workshop for ‘‘source to sink’’ Sedimentary Dynamics in catchment scale, 16–20 June, 2003, Sapporo, Hokkaido University, Japan, pp. 187–206. Luzi, D.S., 2000. Long-term inﬂuence of jams and LWD pieces on channel morphology, Carnation Creek, B.C. unpublished M.Sc., Department of Forest Resources Management, The University of British Columbia, Vancouver, B.C., 148pp.

496

M. A. Hassan et al.

Madej, M.A., 1987. Residence times of channel-stored sediment in Redwood Creek, northwestern California. In: R.L. Beschta, R.L. Blinn, T., Grant, G.E., Ice, G.G., Swanson, F.J. (Eds), Erosion and Sedimentation in the Paciﬁc Rim. Corvallis, Oregon. August 1987. International Association of Hydrological Sciences Publication No. 165. pp. 429–438. Madej, M.A., Ozaki, V., 1996. Channel response to sediment wave propagation and movement, Redwood Creek, California, USA. Earth Surf. Process. Landf. 21, 911–927. Parker, G., Klingeman, P.C., McLean, D.L., 1982. Bedload and size distribution in paved gravel-bed streams. Am. Soc. Civ. Eng. J. Hydraul. Div. 108, 544–571. Parker, G., Toro-Escobar, C.M., Ramey, M., Beck, S., 2003. The effect of ﬂoodwater extraction on the morphology of mountain streams. J. Hydraul. Eng. 129, 885–895. Powell, L.H., 1987. Stream channel morphology changes since logging. In: Chamberlin, T.W., 1987 (Ed.), Proceedings of the workshop: Applying 15 years of Carnation Creek results. Carnation Creek Steering Committee, Paciﬁc Biological Station, Nanaimo, British Columbia, pp. 16–25. Reid, I., Laronne, J.B., 1995. Bedload sediment transport in an ephemeral stream and a comparison with seasonal and perennial counterparts. Water Resour. Res. 31, 773–781. Reid, L.M., Dunne, T., 2003. Sediment budgets as an organizing framework in ﬂuvial geomorphology. In: Kondolf, G.M. and Piegay, H. (Eds), Tools in Fluvial Morphology. Wiley, Chichester, UK, pp. 463–500. Ryan, S.E., 2001. The Inﬂuence of sediment supply on rates of bed load transport: a case study of three streams on the San Juan National Forest. Proceeding of the Seventh Federal Interagency Sedimentation Conference, March 25–29, 2001, Reno, Nevada, pp. III-48–III-54. Ryder, J.M., Fletcher, W.K., 1991. Exploration geochemistry – sediment supply to Harris Creek. British Columbia Min. Energy, Mines and Resour., Geol. Surv. B.C., Geol. Fieldwork, Paper 1991–1, pp. 301–306. Schick, A.P., Lekach, J., Hassan, M.A., 1987. Vertical exchange of coarse bedload in desert streams. In: Frostick, L.E. and Reid, I. (Eds), Desert Sediments: Ancient and Modern, Geological Society of London, Special Publication 35, pp. 7–16. Smith, B., 2004. Relations between bed material transport and storage during aggradation and degradation in gravel bed channel. Unpublished Master thesis, Department of Environmental Systems, Humboldt State University, 108pp. Smith, R.D., 1990. Streamﬂow and bedload transport in an obstruction-affected, gravel-bed stream. PhD. Thesis, Oregon State University, Corvallis, 181pp. Sutherland, D.G., Ball, M.H., Hilton, S.J., Lisle, T.E., 2002. Evolution of a landslide-induced sediment wave in the Navarro River, California. Geol. Soc. Am. Bull. 114, 1036–1048. Swanson, F.J., Fredrickson, R.L., McCorison, F.M., 1982a. Material transfer in a Western Oregon forested watershed. In: Edmonds, R.L. (Ed.), Analysis of Coniferous Forest Ecosystems in the Western United States. Hutchinson Ross Publishing Co., Stroudsburg, PA, pp. 233–266. Swanson, F.J., Janda, R.J., Dunne, T., 1982b. Summary: Sediment budgets and routing in forested drainage basins. In: Swanson, F.J., Janda, R.J., Dunne, T. and Swanston, D.N. (Eds), Sediment Budgets and Routing in Forested Drainage Basins. United States Department of Agriculture, General Technical Report Number PNW-141, pp. 157–165. Tassone, B.L., 1987. Sediment loads from 1973 to 1984 08HB048 Carnation Creek at the mouth, British Columbia. In: Chamberlin, T.W. (Ed.), Proceedings of the workshop: Applying 15 years of Carnation Creek results. Carnation Creek Steering Committee, Paciﬁc Biological Station, Nanaimo, British Columbia, pp. 46–58. Wilcock, P.R., McArdell, B.W., 1993. Surface-based fractional transport rates: Mobilization thresholds and partial transport of a sand-gravel sediment. Water Resour. Res. 29, 1297–1312.

Discussion by Ian Reid Hassan et al. draw attention to the effects of sediment supply on the degree of armour development in gravel-bed channels, using this as an explanation for the differences between the armoured beds of Harris Creek (sediment starved) and

496

M. A. Hassan et al.

Madej, M.A., 1987. Residence times of channel-stored sediment in Redwood Creek, northwestern California. In: R.L. Beschta, R.L. Blinn, T., Grant, G.E., Ice, G.G., Swanson, F.J. (Eds), Erosion and Sedimentation in the Paciﬁc Rim. Corvallis, Oregon. August 1987. International Association of Hydrological Sciences Publication No. 165. pp. 429–438. Madej, M.A., Ozaki, V., 1996. Channel response to sediment wave propagation and movement, Redwood Creek, California, USA. Earth Surf. Process. Landf. 21, 911–927. Parker, G., Klingeman, P.C., McLean, D.L., 1982. Bedload and size distribution in paved gravel-bed streams. Am. Soc. Civ. Eng. J. Hydraul. Div. 108, 544–571. Parker, G., Toro-Escobar, C.M., Ramey, M., Beck, S., 2003. The effect of ﬂoodwater extraction on the morphology of mountain streams. J. Hydraul. Eng. 129, 885–895. Powell, L.H., 1987. Stream channel morphology changes since logging. In: Chamberlin, T.W., 1987 (Ed.), Proceedings of the workshop: Applying 15 years of Carnation Creek results. Carnation Creek Steering Committee, Paciﬁc Biological Station, Nanaimo, British Columbia, pp. 16–25. Reid, I., Laronne, J.B., 1995. Bedload sediment transport in an ephemeral stream and a comparison with seasonal and perennial counterparts. Water Resour. Res. 31, 773–781. Reid, L.M., Dunne, T., 2003. Sediment budgets as an organizing framework in ﬂuvial geomorphology. In: Kondolf, G.M. and Piegay, H. (Eds), Tools in Fluvial Morphology. Wiley, Chichester, UK, pp. 463–500. Ryan, S.E., 2001. The Inﬂuence of sediment supply on rates of bed load transport: a case study of three streams on the San Juan National Forest. Proceeding of the Seventh Federal Interagency Sedimentation Conference, March 25–29, 2001, Reno, Nevada, pp. III-48–III-54. Ryder, J.M., Fletcher, W.K., 1991. Exploration geochemistry – sediment supply to Harris Creek. British Columbia Min. Energy, Mines and Resour., Geol. Surv. B.C., Geol. Fieldwork, Paper 1991–1, pp. 301–306. Schick, A.P., Lekach, J., Hassan, M.A., 1987. Vertical exchange of coarse bedload in desert streams. In: Frostick, L.E. and Reid, I. (Eds), Desert Sediments: Ancient and Modern, Geological Society of London, Special Publication 35, pp. 7–16. Smith, B., 2004. Relations between bed material transport and storage during aggradation and degradation in gravel bed channel. Unpublished Master thesis, Department of Environmental Systems, Humboldt State University, 108pp. Smith, R.D., 1990. Streamﬂow and bedload transport in an obstruction-affected, gravel-bed stream. PhD. Thesis, Oregon State University, Corvallis, 181pp. Sutherland, D.G., Ball, M.H., Hilton, S.J., Lisle, T.E., 2002. Evolution of a landslide-induced sediment wave in the Navarro River, California. Geol. Soc. Am. Bull. 114, 1036–1048. Swanson, F.J., Fredrickson, R.L., McCorison, F.M., 1982a. Material transfer in a Western Oregon forested watershed. In: Edmonds, R.L. (Ed.), Analysis of Coniferous Forest Ecosystems in the Western United States. Hutchinson Ross Publishing Co., Stroudsburg, PA, pp. 233–266. Swanson, F.J., Janda, R.J., Dunne, T., 1982b. Summary: Sediment budgets and routing in forested drainage basins. In: Swanson, F.J., Janda, R.J., Dunne, T. and Swanston, D.N. (Eds), Sediment Budgets and Routing in Forested Drainage Basins. United States Department of Agriculture, General Technical Report Number PNW-141, pp. 157–165. Tassone, B.L., 1987. Sediment loads from 1973 to 1984 08HB048 Carnation Creek at the mouth, British Columbia. In: Chamberlin, T.W. (Ed.), Proceedings of the workshop: Applying 15 years of Carnation Creek results. Carnation Creek Steering Committee, Paciﬁc Biological Station, Nanaimo, British Columbia, pp. 46–58. Wilcock, P.R., McArdell, B.W., 1993. Surface-based fractional transport rates: Mobilization thresholds and partial transport of a sand-gravel sediment. Water Resour. Res. 29, 1297–1312.

Discussion by Ian Reid Hassan et al. draw attention to the effects of sediment supply on the degree of armour development in gravel-bed channels, using this as an explanation for the differences between the armoured beds of Harris Creek (sediment starved) and

Sediment storage and transport in coarse bed streams

497

Carnation Creek (sediment fed). They build a spectrum of channel types where the distinguishing characteristic is the armour ratio and relate this positively with ﬂashiness of the ﬂood hydrograph and inversely with sediment supply. The spectrum is categorized as passing from arid to humid to snow-melt regimes. While they refer to the works of Lisle and Madej (1992), Dietrich et al. (1989) and others which allude to the importance of sediment supply in the development of armouring and, hence, sediment dynamics, they have not mentioned the body of publications by ourselves on the differences in sediment transport between ephemeral and perennial systems and our thesis that ﬂood regime and sediment supply both mean that the former are unarmoured while the latter are armoured. These publications base their argument not only on observations of bed structure, but, signiﬁcantly, on direct measurements of bedload transport. Examples are Laronne and Reid (1993), Laronne et al. (1994) and Reid and Laronne (1995). Readers will usefully be aware of them in order to contextualise Hassan et al.’s paper. Reply by the authors We thank Ian Reid for bringing to our attention some of his work. In the revised text we have included some of his suggested references. References Laronne, J.B., Reid, I., Yitshak, Y., Frostick, L.E., 1994. The non-layering of gravel streambeds under ephemeral ﬂow regimes. J. Hydrol. 159, 353–363. Laronne, J.B., Reid, I., 1993. Very high bedload sediment transport in desert ephemeral rivers. Nature 366, 148–150. Reid, I., Laronne, J.B., 1995. Bedload sediment transport in an ephemeral stream and a comparison with seasonal and perennial counterparts. Water Resour. Res. 31, 773–781.

Sediment storage and transport in coarse bed streams

497

Carnation Creek (sediment fed). They build a spectrum of channel types where the distinguishing characteristic is the armour ratio and relate this positively with ﬂashiness of the ﬂood hydrograph and inversely with sediment supply. The spectrum is categorized as passing from arid to humid to snow-melt regimes. While they refer to the works of Lisle and Madej (1992), Dietrich et al. (1989) and others which allude to the importance of sediment supply in the development of armouring and, hence, sediment dynamics, they have not mentioned the body of publications by ourselves on the differences in sediment transport between ephemeral and perennial systems and our thesis that ﬂood regime and sediment supply both mean that the former are unarmoured while the latter are armoured. These publications base their argument not only on observations of bed structure, but, signiﬁcantly, on direct measurements of bedload transport. Examples are Laronne and Reid (1993), Laronne et al. (1994) and Reid and Laronne (1995). Readers will usefully be aware of them in order to contextualise Hassan et al.’s paper. Reply by the authors We thank Ian Reid for bringing to our attention some of his work. In the revised text we have included some of his suggested references. References Laronne, J.B., Reid, I., Yitshak, Y., Frostick, L.E., 1994. The non-layering of gravel streambeds under ephemeral ﬂow regimes. J. Hydrol. 159, 353–363. Laronne, J.B., Reid, I., 1993. Very high bedload sediment transport in desert ephemeral rivers. Nature 366, 148–150. Reid, I., Laronne, J.B., 1995. Bedload sediment transport in an ephemeral stream and a comparison with seasonal and perennial counterparts. Water Resour. Res. 31, 773–781.

Gravel-Bed Rivers VI: From Process Understanding to River Restoration H. Habersack, H. Pie´gay, M. Rinaldi, Editors r 2008 Elsevier B.V. All rights reserved.

501

19 Ecological responses to anthropogenic alterations of gravel-bed rivers in Japan, from ﬂoodplain river segments to the microhabitat scale: a review Futoshi Nakamura, Yoˆichi Kawaguchi, Daisuke Nakano and Hiroyuki Yamada

Abstract We describe the relationship between disturbance regimes, the life history traits of aquatic and riparian organisms, and effects of human activity, using Japanese gravelbed rivers in the Asia Monsoon Belt as an example. The consideration of various roles of disturbance in creating a spatial and temporal pattern of habitats is made hierarchically at three spatial scales. Segment scale is the largest, represented by a braided river landscape on an alluvial fan. Riparian tree species direct their life history strategies to survive in the shifting habitat dynamics. Reservoirs constructed at the fan apex regulate ﬂood disturbances and seasonal ﬂow variation, changing the dynamic state to a monotonous, static habitat structure. Intermediate reach-scale structures are represented by pool-rifﬂe sequences, gravel bars, secondary and abandoned channels and oxbow lakes. We focused particularly on the lateral variation of the entire valley ﬂoor and its stream channel, which provide critical habitats for spawning, hatching, rearing, wintering, feeding and dwelling as well as ﬂow refugia for ﬁsh and macroinvertebrates. Bed instability resulting from an increased tractive force due to channel straightening and its impact on macroinvertebrate communities is also discussed. Finally, the smallest scale is the microhabitat seen as the interstitial spaces in gravel beds used by salmonid and benthic ﬁsh for all or some of their life stages. Sediment control dams and gravel mining in rivers and ﬂoodplains is causing gravel-size sediment starvation, whereas the introduction of ﬁne sediment is substantially increased following land-use development. 1.

Introduction

Rivers and ﬂoodplains are highly dynamic in space and time, and the indigenous plant and animal species exploit fully their inherited potential to survive in the E-mail address: [email protected] (F. Nakamura) ISSN: 0928-2025

DOI: 10.1016/S0928-2025(07)11141-X

502

F. Nakamura et al.

disturbance-prone areas. Braided rivers with gravel beds are typically observed in alluvial fan systems where valley ﬂoor landforms are relatively unconstrained and reﬂect generally steep bed gradients (Pie´gay et al., 2006). The most common disturbance to the stability of local landform morphology in braided rivers is ﬂooding. During ﬂoods, multiple active channels migrate laterally, ﬁlling abandoned and secondary channels. In general, ﬂood frequency is on the order of years within active channels, and ten to a hundred years in the more stable ﬂoodplains (Nakamura, 1990; Nakamura et al., 1995). However, lateral channel migration is able to alter stable ﬂoodplains subject to frequent and intensive ﬂood events into highly unstable areas. In Asian Monsoon areas, heavy seasonal rainfalls cause ﬂooding in both monsoons (June to early July) and typhoons (August to early October). In the northern part of Japan, high ﬂows resulting from snowmelt are also signiﬁcant, and these occur each spring (April and May). When ﬂooding completely changes the surface topography of a braided river, the geomorphic conﬁguration is changed such that new surfaces with fresh mineral grains are exposed during the periods of water recession. Sediment, nutrients, and woody debris transported from upstream accumulate to form complex geomorphic surfaces that have a variety of substrate properties and nutrient conditions (Dollar, 2000). Bed materials that consist of pebble- and gravel-size particles are porous and contribute to the creation of a dynamic and extensive hyporheic zone (Wondzell and Swanson, 1999). Plant and animal species are adapted to highly unstable ﬂuvial environments. For instance, new exposed geomorphic surfaces created by ﬂooding are typically recruitment sites for pioneer wind-dispersed and light-demanding species (Nakamura et al., 1997). Secondary and abandoned channels as well as backwaters provide refuge for Oncorhynchus masou (masu salmon) during unpredictable ﬂooding associated with monsoons or typhoons. The hyporheic zone created within mobile, porous gravel beds not only plays a role in water quality but can also provides a living and spawning habitat for ﬁsh and benthic invertebrates (Stanford and Ward, 1993). In Japan and other East Asian countries, rivers in alluvial fans have been intensively altered during urban and agricultural development after World War II. Dam construction and river channelization generally lower the frequency of highmagnitude ﬂood disturbance and increase low-magnitude disturbance, resulting in reduction of habitat diversity. For instance, reservoirs built for power generation and ﬂood control at the alluvial fan apex of a braided river regulate water discharge, thus diminishing the seasonal variation of ﬂood disturbance. Dikes and revetments in the river channel constrain the lateral mobility of river courses, changing the braided morphology into a narrow meandering course characterized by alternating gravel bars. Furthermore, the beds of many gravel-bed rivers have been degraded and armored following gravel mining, channelization and dam construction during the 1960s and 1970s. Additionally, the introduction of ﬁne sediment produced as a result of forestry and agricultural activities has been increased over much of the area since 1970s, and this has plugged interstitial spaces in the gravel beds and diminished the spatial extent and quality of the hyporheic zone. The objective of this chapter is to describe the relationship between ﬂood disturbance regimes that create a mosaic of shifting habitat and life history traits for

Ecological responses to anthropogenic alterations of gravel-bed rivers

503

dominant tree, ﬁsh and benthic macroinvertebrates species, using the gravel-bed rivers of Japan as an example. Moreover, we wish to show how plant and animal species that depend on ﬂood disturbance have declined through the loss of habitat diversity following the alteration of the river morphology and regulation of the river ﬂow. Our discussion is based on three spatial scales that are hierarchically organized in scale, as follows. Firstly, segment scale is the greatest, represented by an entire stretch (the order of kilometers) of gravel-bed river within an alluvial fan. Riparian tree species have adapted to surviving in this disturbance regime. The resulting landscape at the segment scale can be characterized as a ‘‘shifting habitat mosaic’’. Here, the shifting habitat mosaic refers to the mosaic of ﬂoodplain habitat patches, which are continuously changing and in a phase of successional development (Bormann and Likens, 1979). Human impacts that greatly affect segment scale river dynamics include hydropower and ﬂood control reservoirs. Secondly, there are reach-scale structures represented by pool-rifﬂe sequences, gravel bars, secondary and abandoned channels and oxbow lakes. The foci of our discussion at this scale are the lateral variations of a river valley ﬂoor and its main channel, with special reference to habitat diversity for ﬁsh and macroinvertebrates. The role of large woody debris within this scale is also discussed. Since the lateral variation of river morphology and its associated habitat diversity have been lost from within Japanese gravel-bed river systems following dike construction and channelization, the bed instability created by the increased tractive force of water ﬂow due to channel straightening and its impact on invertebrate communities is also considered. Finally, the smallest scale is the microhabitat, which we discuss with relation to the interstitial voids in gravel beds that are used by salmonid and benthic ﬁsh for all or a portion of their life-stages. The groundwater regime in a gravel-bed river provides a complex and extensive hyporheic habitat that is, however, being destroyed at an alarming rate in Japan. Sediment control-check dams in catchment headwaters are causing gravel-size alluvial sediment starvation downstream, whereas ﬁne sediment production has been substantially increased following land-use development.

2. 2.1.

Dynamic states of a braided river: shifting habitat mosaic in segment scale Riparian forest dynamics

A typical disturbance to a braided river is large-scale ﬂooding, which can extend completely across the ﬂoodplain and can form various geomorphic surfaces (e.g., gravel bars, low- and high-ﬂoodplains, secondary channels) that provide regeneration habitats for riparian tree species (e.g., Salix and Populus spp.). Floods frequently occur in braided rivers because of seasonal snowmelt and irregular monsoon and typhoon high ﬂows. Because of the frequency of ﬂooding that periodically creates new recruitment habitats, pioneer species that have a seedling regeneration and a short life span are clearly among the favored plant varieties to occur in braided river valleys.

504

F. Nakamura et al.

Riparian forests consisting of pioneer species represented by Salicaceae grow in the gravel-bed river valleys of Hokkaido, Japan (Niiyama, 1990) and in other locations throughout the northern temperate zone (Karrenberg et al., 2002). As Osterkamp and Hupp (1984) and Hupp and Osterkamp (1985) pointed out, a close relationship exists between ﬂuvial landforms and disturbance frequencies whereby certain species are signiﬁcantly associated with speciﬁc ﬂuvial landforms. Shin and Nakamura (2005), in a study of the Rekifune River in northern Japan, demonstrated that landforms classiﬁed according to ﬂooding frequency show different environmental gradients that strongly regulate the recruitment sites of riparian species and inﬂuence the adaptation of life history strategies (Fig. 19.1). In their study, ﬂoodwater levels for 2- and 20-year recurrence intervals (RI) were used as thresholds. The gravel bar habitat is the area submerged when the water level rises sufﬁciently high to ﬁll the present mainstream channel. A lower ﬂoodplain habitat, along the mainstream channel, is the area submerged during a 2-year RI ﬂood. A secondary channel habitat, cutting through riparian forests, has the same submergence probability as the lower ﬂoodplain. The upper ﬂoodplain habitat is the area that submerges with a 20-year RI ﬂood. The terrace habitat is not submerged even at this recurrence interval. Chosenia arbutifolia, Toisusu urbaniana and Populus maximowiczii dominate gravel bars, and the lower and upper ﬂoodplains. These geomorphic surfaces, characterized by gravelly soil and low soil moisture availability, are very prohibitive recruitment sites for other riparian tree species, but the mentioned species survive (Niiyama, 1987, 1989). The axial roots of these plants grow quickly in a coarse gravel soil, particularly during the early growth stages (Ishikawa, 1994). This fast-growing root system enables these species to absorb water from deeper sources, and permits them to resist physical damage and erosion. They can easily establish themselves on gravel bars where water levels change constantly. The maturation age of C. arbutifolia and T. urbaniana ranges from 10 to 15 years, and the ages of the oldest individuals in the Rekifune River valley were found to be between 35 and 40 years (unpublished data). The turnover time of the entire active ﬂoodplains by ﬂooding for the Rekifune River was about 30 years (Nakamura and Shin, 2001), indicating that the short life span of the oldest plants nearly synchronizes with this ﬂood recurrence. Young maturation is one of the ﬂoral life history strategies required to maintain a successful recruitment to areas subject to periodic ﬂoods. Salix pet-susu and Salix sachalinensis are resistant to prolonged inundation. An in situ experiment demonstrated a 100% survival rate after these plants had been submerged for more than two months (Nagasaka, 1996). Such tolerance to waterlogging enables these species to survive in areas of poorly drained soil and anoxic conditions that occur in swamps (Niiyama, 1995). Such wet environments characterized by a silty sand substrate and low pF (soil moisture tension) conditions are favored by the two species. Salicaceae is particularly dominant in the active gravel-bed rivers of Japan. The extremely light seeds (38–600 mg for 1,000 seeds) are produced proliﬁcally each year and are easily dispersed by wind over a long distance (Sato, 1995; Niiyama, 2002). Seed germination requires an exposed surface with a mineral soil of sandy pebbles, as is formed extensively by ﬂooding. Seeds are released during the period of descending

Ecological responses to anthropogenic alterations of gravel-bed rivers

505

+1.0

Carbon Nitrogen

UF7

UF4

pF value

A-layer

UF1 UF9 UF8

UF6

TE2

UF2

LF2

UF10

Tu

Gravel

Pm

GB7 GB9 LF7

UF3

GB4 GB6 LF3 Ca GB8 LF5 GB1 LF4 LF6 LF8 LF1 LF10 LF9 SC8 GB3 SC7 SC5

TE9 TE5

UF5

Sand

Uj

Silt-clay Ah

SC9 GB5

TE4

Fm

TE6 SC1

TE10 TE7 TE1

SsSp

GB10

Open circle codes :

TE8 TE3

SC4 SC2 GB2

SC3

SC6 SC10

GB: Gravel bar LF: Lower floodplain SC: Secondary channel UF: Upper floodplain TE: Terrace

-1.0

Braided: 1-5 Meandering: 6-10 -1.0

+1.0

Figure 19.1. Biplot graph of Canonical Correspondence Analysis indicating relationships among geomorphic surfaces, environmental variables and dominant species in the Rekifune River, Japan. Arrow directions indicate signiﬁcance, lengths being proportional to importance. Numbers from 1 to 5 and from 6 to 10 with open circles indicate quadrats in the braided and incised meandering section, respectively. Solid circles indicate distribution of dominant species: ‘Ca’: Chosenia arbutifolia; ‘Tu’: Toisusu urbaniana; ‘Pm’: Populus maximowiczii; ‘Ss’: Salix sachalinensis; ‘Sp’: Salix pet-susu; ‘Ah’: Alnus hirsuta; ‘Uj’: Ulmus japonica; ‘Fm’: Fraxinus mandshurica var. japonica. The geomorphic surfaces are classiﬁed according to ﬂood frequencies. (From Shin and Nakamura, 2005, permission for reproduction obtained from Springer Netherlands.)

limb of hydrograph between April and September in snowmelt ﬂoods, with timing differing slightly between the Salicaceae species (Niiyama, 1990; Nagasaka, 1996). Salix rorida starts seed dispersal in mid-May, whereas Salix subfragilis waits until June to early July. The differences in dispersal times by genus are quite distinct. Seed dispersal for C. arbutifolia and P. maximowiczii occurs from late June to early July, which is later than that of Salix spp. T. urbaniana is even later, between August and

506

F. Nakamura et al.

Figure 19.2. Shifting habitat mosaic in the Rekifune River where natural ﬂow regime is preserved.

September. As successive gravel bar surfaces are exposed by the descending ﬂoodwater, they are occupied progressively by the early and late seed-dispersal species. The latest species may colonize areas close to the channel thalweg. In order to complete the entire life cycle of these pioneer species, however, some of these recruitment sites need to be stable long enough for species maturation, such as the more stable surfaces on lower and higher ﬂoodplains. Moreover, regeneration habitats from the germination to maturation stage need to be created continuously in the ﬂoodplain and maintained dynamically by ﬂuvial geomorphic disturbances to generate a shifting habitat mosaic (Fig. 19.2). Seedlings of the shade-tolerant species, Ulmus davidiana var. japonica and Fraxinus mandshurica var. japonica, are less light demanding and are able to colonize areas beneath mature Salicaceae trees (Nanson and Beach, 1977; Shin et al., 1999). They become established on terraced surfaces characterized by a high organic content and deep humus rich soil. However, the conditions for germination under canopy trees are restricted by shading, disease and herbivores, or in some cases by desiccation from being exposed on leaf litter. Successful seedling establishment on ﬂoodplain terraces is more likely to occur on exposed mineral soils. These are revealed during ﬂood disturbances that eliminate forest ﬂoor vegetation (Seiwa, 1997). Because of the relatively high elevation of terraces, even large-scale ﬂooding may disturb only the forest ﬂoor and may leave canopy trees with very little damage. Successful recruitment for these species therefore requires both a long life span for the mature trees, in order for them to continuously disperse seeds to unpredictable regeneration sites, and a shadetolerant character, to enable them to grow under canopy trees. Species that have such life history traits will grow slowly under mature Salicaceae trees but will have a longer life span. For example, U. davidiana var. japonica can live for about 200–300 years (Kon and Okitsu, 1999) in contrast to the short life spans of C. arbutifolia (about 100 years), P. maximowiczii (about 110 years) and T. urbaniana (about 80 years) (Shin et al., 1999). When mature Salicaceae trees die of senescence, shade tolerant, mid-successional species can be released from suppression and grow rapidly.

Ecological responses to anthropogenic alterations of gravel-bed rivers 2.2.

507

Effects of river regulation on regeneration processes of riparian trees

In recent years, however, mature riparian Salicaceae forests in Japanese gravel-bed rivers have been disappearing at an accelerated rate because of land-use effects and river regulation caused by developments such as dikes and channelization associated with urbanization and reservoir construction. For example, rare pioneer species represented by C. arbutifolia have been replaced by shade-tolerant species. It is therefore insufﬁcient merely to conserve existing stands (Johnson et al., 1976; Shin et al., 1999; Gilvear et al., 2000). Rather, the conditions of the shifting habitat mosaic and the river dynamics maintaining those conditions need to be preserved to conserve the pioneer species (Shin and Nakamura, 2005). Reservoirs, erosion control dams and similar dam structures are thought to have considerable effects on the structure, composition and regeneration processes of riparian forests by regulating river ﬂows and trapping sediment (e.g., Reily and Johnson, 1982; Harris et al., 1987; Rood and Heinze-Milne, 1989). The effects of such structures on tree regeneration may change depending on the life history stages of each species. A large reservoir was constructed in 1997 for ﬂood control, water diversion and power generation in the Satsunai River, near the Rekifune River. Since then, the pattern of ﬂood disturbance and seasonal ﬂow variation has altered dramatically. Based on the results of direct gradient analyses, Nakamura and Shin (2001) and Takagi and Nakamura (2003) concluded that a diminished ﬂood frequency should alter the distribution of riparian tree species in Satsunai River. The stands of pioneer species represented by C. arbutifolia, T. urbaniana and P. maximowiczii are receding without regenerating, while shade-tolerant species such as U. davidiana var. japonica and F. mandshurica var. japonica are gradually dominating these areas. A predictive simulation by Nakamura and Shin (2001) indicated that regeneration sites for these pioneer species will disappear in 25 years; thus, C. arbutifolia and T. urbaniana will not be able to complete their life cycle and may eventually become locally extinct. There are many reports that woodland expansion progresses in regulated river valleys where there has been a decrease in the exposure of bare gravel surfaces. Harris et al. (1987) concluded that woodland expansion following dam construction was a result of a loss of scoured areas due to the reduction of peak ﬂow. Johnson (1994) also reported, in his study in Nebraska, U.S.A, that the most advanced woodland expansion since 1900 eliminated sandbars at an annual rate of 10%. He also noted that forest expansion progressed toward the lower riverine reaches. One of the reasons for this woodland expansion in Nebraska was the decrease in river ﬂow rates in June due to water storage in dams and its use for irrigation. The reduced water levels provided recruitment sites that would have been submerged by preregulated water levels. The tendency toward woodland expansion has also been reported in Japan in recent years (Lee et al., 1998). Based on our predictions (Nakamura and Shin, 2001; Takagi and Nakamura, 2003), the gravel-bar-dominant river landscape will be greatly altered by forest expansion as a result of ﬂow regulation, with the dominant trees shifting from pioneer to late successional species. The latter, which are normally predominant on hill slopes, will be eventually established in the former ﬂoodplains when these are deprived of river dynamics.

508 3.

F. Nakamura et al. Lateral variation of river morphology and habitat diversity: reach scale

3.1. Habitat diversity across the river-floodplain system and life history traits of fish and invertebrate In a naturally ﬂowing river valley, various aquatic habitats within a wide range of ﬂow conditions are formed across the river-ﬂoodplain system, from the main channel and side arms, to backwaters and ﬂoodplain pools, to oxbow lakes and swamps (Ward et al., 1998). These habitat types differ in physical and chemical properties, thereby each supporting a characteristic biological community that specializes to adjust to the different abiotic properties. Running and standing water habitats in particular have traditionally been considered as different ecosystems because of large distinctions in their community structures and compositions (Wallace and Anderson, 1996; Wetzel, 2001). Although many ﬁsh use a main channel of ﬂowing water for temporal migration, spawning and shelter from drought, overall ﬁsh biomass across a ﬂoodplain depends on the spatial extent of semiﬂowing water such as backwaters and ﬂoodplain pools connected to the main channel, and standing water temporarily connected during high ﬂow (e.g., oxbow lakes) (Junk et al., 1989). It has also been reported that community compositions of both ﬁsh and benthic macroinvertebrates differ markedly between semiﬂowing waters and standing waters (Battle and Mihuc, 2000; Buffagni et al., 2000; Kawaguchi et al., 2005; Nakano et al., 2005). Furthermore, there are physiochemical and biological differences between main and side channels, with more rheophilic benthic species found in the main channels (Marchese and Ezcurra de Drago, 1992). A river-ﬂoodplain system thus exhibits a cross-sectional variation of river morphology, which comprises a range of aquatic habitats with various physical and chemical environments, thereby increasing the diversity of community composition and its productivity. Flood disturbance and lateral channel migration play important roles in creating, removing and connecting the ﬂoodplain aquatic habitats, which is vital to the maintenance of biodiversity at the entire ﬂoodplain scale (Robinson et al., 2002; Ward et al., 2002). The life history traits of many ﬁsh species in a large river depend on the shifting habitat mosaic, which is maintained by a dynamic interaction between the main channel, secondary channels and the ﬂoodplain. Many ﬁsh species change their habitats, responding to elevating water levels due to ﬂooding (Robinson et al., 2002). Ephemeral semiﬂowing water (ﬂoodplain pools connected temporarily with their main stem), permanent semiﬂowing water (backwaters) and standing water (oxbow lakes) constitute an essential river-ﬂoodplain habitat that allows in-stream species to complete their life cycle (Halyk and Balon, 1983; Junk et al., 1989; Scott and Nielsen, 1989; Penczak et al., 2003). It has been found that ﬁsh species inhabiting the main stem of a river use its ephemeral ﬂoodplain pools as rearing sites for juvenile ﬁsh and for shelter during high-ﬂow events (Halyk and Balon, 1983). They stressed the importance of hydrologic connectivity between a main channel and its ﬂoodplain pools for ﬁsh productivity across the whole ﬂoodplain. Compared to main channels, backwaters and oxbow lakes are characterized by their relatively slow or nearly standing ﬂow conditions. Studies have reported that

Ecological responses to anthropogenic alterations of gravel-bed rivers

509

backwaters can serve as spawning sites for those ﬁsh species preferring slow water (Junk et al., 1989; Scott and Nielsen, 1989). Moreover, oxbow lakes can be important spawning and rearing sites for juveniles (Penczak et al., 2003). Newly hatched larvae and fry of Cyprinidae and many other species use extremely slow waters (Houde, 1969; Harvey, 1987; Sagawa et al., 2005). This indicates that semiﬂowing and standing waters, including ﬂoodplain pools and backwaters connected to a main channel and oxbow lakes, provide critical rearing habitats for many ﬁshes, especially for young ﬁsh. Many ﬁsh species thus move back and forth in their life cycle along a gradient of ﬂow conditions that vary from ﬂowing (i.e., main channel) to standing waters (i.e., riverine wetlands and oxbow lakes). In contrast, most benthic macroinvertebrate species are adapted to either ﬂowing or standing water. They stay in either a lotic or a lentic habitat throughout the in-stream stages of their life cycles (Wallace and Anderson, 1996; Smock, 1999). However, some species of Siphlonisca, Siphlonurus and Leptophlebia are adapted to highly predictable ﬂuvial disturbances such as a snowmelt ﬂood. They move from the main channels to the inundated ﬂoodplains to reach standing water temporarily created in spring and summer (Huryn and Gibbs, 1999). A mayﬂy species, Siphlonisca aerodromia is one of the river-ﬂoodplain fauna that uses both main channels and the ﬂoodplain habitats during its life cycle; S. aerodromia grows rapidly in the ephemeral standing water and emerges afterward. The adult of this species lays eggs in ﬂowing water to avoid damage to eggs and young by drought, or cold during winter (Gibbs and Mingo, 1986; Gibbs and Siebenmann, 1996). Macroinvertebrate species dwelling in permanent ﬂowing water can be washed away to perish during high ﬂow following unpredictable rain storms (Resh et al., 1988; Robinson et al., 2003; Robinson et al., 2004). Poff and Ward (1989) identiﬁed adaptations to irregular but frequent disturbances including a behavioral response avoiding disturbances, as well as small body size, rapid growth rate and intraspecies asynchronous larvae development. For example, some species of mayﬂy nymphs dwelling in rifﬂes in base-ﬂow conditions migrate into slow current backwaters to avoid being swept away in ﬂoods (Lehmkuhl and Anderson, 1972). Lancaster and Hildrew (1993) described in-stream refugia in which dragging force and ﬂow velocity increased little during high-ﬂow events. The known in-stream ﬂow refugia during high ﬂow for benthic organisms are large wood and stable stones (Borchardt, 1993; Palmer et al., 1996; Matthaei et al., 2000), backwaters and temporarily submerged stream banks (Badri et al., 1987; Gladden and Smock, 1990; Matthaei and Townsend, 2000; Negishi et al., 2002). Presence of these features suggest that the lateral variation in river morphology with its full range of aquatic habitats and ﬂow conditions can minimize the loss of benthic biota in high-ﬂow events and is effective in accelerating the postdisturbance recovery of the community.

3.2.

Lateral variation of habitat across the main channel

Gravel-bed rivers typically display a longitudinal alternation of rifﬂe-pool sequences at the reach scale. Abundant literature has characterized the physical condition differences between rifﬂes and pools and has related these characteristics to the

510

F. Nakamura et al.

distribution of in-stream biota (Brown and Brussock, 1991; Inoue and Nakano, 1994; Inoue et al., 1997; Inoue and Nakano, 1999). Lateral variation in the physical environment can also be found across the width of a river, although the variation is not as large as it is across its ﬂoodplain, which ranges from ﬂowing to standing water. Near-shore zones (the edge habitat) are shallower, have slower currents than those in the thalweg (Negishi et al., 2002; Nakano and Nakamura, 2006), and accumulate a large amount of organic matter (Brewin et al., 1995). The edge and thalweg habitats also differ in the community composition of benthic species because of the different physical conditions and associated substrate development (Ormerod et al., 1993; Brewin et al., 1995). Fish segregate their habitats depending on their species, and life stage within the same species, and therefore habitat types with different physical conditions differ in ﬁsh composition and abundance (Modde et al., 1991; Inoue and Nakano, 1999). An individual species requires a variety of habitat types during its life cycle. Of these, the rearing habitat for young-of-the-year class, including larvae and fry that have high mortality rates, is often thought to be very important. Young-of-the-year ﬁsh selectively use slow water within a main channel, such as topographic depressions or near-shore areas covered by overhanging vegetation (the lateral habitat), to avoid fast ﬂow in the mid-channel areas (Sagawa et al., 2005). The channel cross-section at a river bend is asymmetrical because of differential current velocity, with a deep scour pool on the outside of the curve created by erosion, and a shallow area over a point bar formed by deposition on the inside of the bend (Morisawa, 1985). This lateral variation in the physical environment of a main channel may be expected to promote species diversity in macroinvertebrates. Studies in the lower reaches of the gravel-bed Nishibetsu River in Japan found that macroinvertebrates concentrated in the shallow areas inside the bends, yet they were sparse in the deep water on the outside, with extremely low density and low species richness (Nakano et al., 2005; Nakano and Nakamura, 2006). Similarly, a study in the large Canadian gravel-bed Fraser River, reported that macroinvertebrate communities were poorly distributed in its deepest part (Rempel et al., 2000). Community responses to the variation of physical habitat across the main channel of a gravel-bed river are thus different from those across the entire ﬂoodplain of the same river. The diversity of in-stream biota increases with increasing diversity of physical environment in the latter cases, whereas such a biological gradient does not occur in the former. This is probably because of low riverbed stability in deep water. Generally, as water depth increases, shear stress on the riverbed increases, and bed material can be transported. Substrate materials in both rivers consist of small clasts with an average diameter of approximately 10 mm in the Nishibetsu River (Nakano et al., 2005) and 15–35 mm in the Fraser River (Rempel et al., 1999), and are therefore presumably highly mobile under normal ﬂow conditions. There was a negative relationship between shear stress and the density and number of macroinvertebrates in the Nishibetsu (Fig. 19.3) and Fraser rivers. In rivers with small-diameter riverbed material, the shallow areas along the shores created by point bars or gravel deposits become an important edge habitat for benthic macroinvertebrates. Similarly for ﬁsh species, the area on the outside of a channel bend is less likely to be a favored habitat, because of the concentrated ﬂow and fast currents (Kawaguchi

Ecological responses to anthropogenic alterations of gravel-bed rivers 30

8000 (a)

(b) Taxon richness (no./sample)

7000 Total density (ind./m2)

511

6000 5000 4000 3000 2000 1000

25 20 15 10 5 0

0 3.0

4.0

5.0

6.0

7.0

8.0

9.0

10.0

3.0

4.0

5.0

6.0

7.0

8.0

9.0

10.0

Shear velocity (cm/s) Figure 19.3. Total macroinvertebrate density (a), and taxon richness (b) in relation to shear velocity in the Nishibetsu River, Japan. The samples were taken ﬁve times (June and November between 2002 and 2004) across one cross-section of meander bend.

et al., 2005). However, bank erosion by the accelerated ﬂow on outside bends can introduce large woody debris that can mediate high ﬂow velocities and thereby provide ﬂow refugia. In the Shibetsu River in Japan, where restoration of the stream meanders and ﬂoodplain is underway, experimental remeandering work promoted bank erosion on the outside of a bend, and trees fell into the stream. (Fig. 19.4a). The downed trees created localized slow water on the outside of the bend. A diving survey found that large-size of O. masou (masu salmon) migrating upstream used this habitat more frequently than the straight channel that existed before the restoration. Further diving and cast-net surveys conﬁrmed that larger numbers of young O. masou colonized the remeandered reach rather than a controlled straight reach (Fig. 19.4b; Kawaguchi et al., 2005). In-stream cover provided by large wood can become an important habitat for many salmonid (Inoue and Nakano, 1998; Urabe and Nakano, 1998; Abe and Nakamura, 1999). Large woody debris provides a range of important functions for stream salmonids and benthic macroinvertebrates, providing habitat (Nilsen and Larimore, 1973; Benke et al., 1984), food resources (Hoffmann and Hering, 2000) and shelter from high ﬂow (Borchardt, 1993). The importance of these ecological functions is particularly enhanced in rivers with unstable bed materials (Benke and Wallace, 2003). A lateral variation in geomorphic features and ﬂow conditions can be developed within a main channel in association with large woody debris, and this enhances the diversity of the biotic community.

3.3.

Effects of stream channelization on fish and invertebrate

Stream channelization for ﬂood control diminishes the inundation frequency of a ﬂoodplain by isolating the aquatic habitats distributed across it or by reducing

F. Nakamura et al.

512

(a)

Fish abundance

30

20

10

0

(b)

Straight channel

Re-meandering channel

Figure 19.4. (a) Photograph of the large woody debris in the remeandering channel. (b) Fish abundance (mean71SE, n ¼ 4) counted by direct observation and cast net after remeandering in Shibetsu River, Japan. The straight channel is a control reach sampled in August, 2002 (post-experiment), the same time as the remeandering reach. (From Kawaguchi et al., 2005, permission for reproduction obtained from Ecology and Civil Engineering Society, Japan.)

Ecological responses to anthropogenic alterations of gravel-bed rivers

513

connectivity among them. Reduced inundation frequency also leads to the disruption of the geomorphic processes that create ﬂoodplain habitats, thereby eliminating lateral habitat variation. This in turn causes signiﬁcant damage to the species that have life history traits adapted to the ambient ﬂood regime and that utilize the hydrological connectivity of the ﬂoodplain habitats. The channelization may affect the organization of an entire river ecosystem (Michener and Haeuber, 1998; Penczak et al., 2000; Winemiller et al., 2000). Floodplain habitat fragmentation can have great impacts, particularly on the young-of-the-year ﬁsh populations of many species (Hohausova´ et al., 2001). Habitat fragmentation also threatens the life cycle of benthic macroinvertebrates that depend on both the main channel and its ﬂoodplain habitat (Huryn and Gibbs, 1999). Stream channelization reduces in-stream habitat diversity by simplifying the channel morphology and in-stream physical conditions (Brookes, 1992). Many studies have determined that the loss of rifﬂe-pool alternation by channelization reduces the longitudinal habitat variation along its stream length, which leads to decreased density and species richness of ﬁsh and benthic invertebrates (Edwards et al., 1984; Takahashi and Higashi, 1984; Wilcock and Essery, 1991; Shimatani et al., 1994; Toyoshima et al., 1996). For example, in a channelized river in Montana, U.S.A., the loss of rifﬂe-pool structures eliminated the deep, slow water and increased the shallow, fast water, resulting in the simpliﬁcation of ﬁsh community structures (Elser, 1968). A study in Japan reported that the rifﬂe-pool sequence was less apparent in a channelized reach than in an intact one. The ﬁsh community structure in the channelized reach was simpler, with the dominance of a single benthic species and a low density of ﬂuvial ﬁsh (Takahashi and Higashi, 1984; Inoue and Nakano, 1994). Moore and Gregory (1988) demonstrated the importance of the cross-sectional habitat diversity for salmon fry. Stream channelization greatly diminishes not only the longitudinal diversity of in-stream habitats but also the lateral diversity. For example, channelization disrupts the formation of stable gravel bars, creating a trapezoidal channel section characterized by homogeneous physical properties across the width of the river (Downs and Thorne, 1998). The number of benthic macroinvertebrate species was greatly reduced in these channelized streams (Nakano et al., 2005). The areas of inundated habitat during high ﬂow, which serve as ﬂow refugia for benthic macroinvertebrates, are also limited in channelized streams (Negishi et al., 2002). Vegetation cover growing in the water along a stream bank reduces the ﬂow velocity near the shore, which serves as an important shelter from predators. In streams where a concrete revetment made during channelization replaces overhanging streamside vegetation, ﬁsh and crustacean abundance can be dramatically reduced (Kawaguchi et al., 2006). Moreover, channel straightening removes riparian trees, and revetment prevents the stream banks from lateral erosion, reducing the delivery of large logs into the stream (Nagasaka and Nakamura, 1999). An experimental removal of logs demonstrated that the volume of pools formed by large woody debris was positively associated with ﬁsh abundance (Abe and Nakamura, 1999). The authors stressed the importance of wood in creating habitats for stream salmonids. Experimental studies have also determined that wood removal reduced the numbers and density of Cyprinidae and Centrarchidae ﬁsh (Angermeier and

F. Nakamura et al.

514

Karr, 1984) and the density and growth of Oncorhynchus kisutch (coho salmon) and Salvelinus malma malma (Dolly Varden) (Dolloff, 1986). Another study reported that wood removal lowered the number of large-size ﬁsh (Angermeier and Karr, 1984; Elliot, 1986).

4. 4.1.

Interstitial space of gravel beds: microhabitat scale Loss of openwork gravel habitat with fine sediment accumulation

The microhabitat structure used by ﬁsh to take shelter from predators and strong currents is termed the ‘‘cover habitat’’ (Shirvell, 1990). Inoue et al. (1997) and Abe and Nakamura (1999) reported that woody debris submerged or overhanging a cut bank provided cover habitat for juvenile O. masou. For benthic ﬁsh, however, interstitial spaces in a gravel bed function as a cover habitat more effectively than do vegetation or microtopography. For example, benthic ﬁsh species such as Cottus nozawae (wrinklehead sculpin) do not migrate a long distance for spawning; rather, they spawn in ordinarily inhabited or adjacent rifﬂes (Goto, 1994). Female adult of C. nozawae lay eggs underneath the gravel bed where male adult set their territory. The male adults guard the eggs until they hatch (Goto, 1993, 1994). Thus, interstitial spaces in a gravel bed are essential habitats in the successful reproduction of many benthic ﬁshes. The ﬁne sediment loading into streams has, however, been increasing because of human activity including land-use development (Walling, 1990; Richards et al., 1993; Allan et al., 1997; Murakami et al., 2001), forestry (Platts et al., 1989), mining (Van Nieuwenhuyse and LaPerriere, 1986; Davies-Colley et al., 1992) and road construction (Barton, 1977; Extence, 1978; Cline et al., 1982). Turbidity and ﬁne sediment have signiﬁcant adverse effects on in-stream biota through the alteration of the physical and chemical environment (Newcombe and MacDonald, 1991; Waters, 1995; Wood and Armitage, 1997). In gravel-bed rivers, a change in substrate properties may have a substantial impact on biotic communities. Murakami et al. (2001) found a negative relationship between the proportion of ﬁne sediment (o2 mm) and the percentage of openwork gravel, as well as the permeability of the bed material as determined using the packer test (Yamada et al., 2005). Turbidity and ﬁne sediment loads in Japanese gravel-bed rivers have increased with the expansion of agricultural areas in their catchments. In general, ﬁne sediment particles accumulate on riverbeds where the shear velocity of surface water is lower than the settling velocity of the particles (Crisp and Carling, 1989). However, under the different hydraulic conditions, ﬁne sediment particles can inﬁltrate into the pores of a riverbed, even if shear velocity exceeds the settling velocity (Beschta and Jackson, 1979; Scha¨lchli, 1992). Inﬁltration of ﬁne sediment causes a decrease in intragravel ﬂow rates (or velocity) with decreasing permeability of the riverbed. The permeability coefﬁcient inversely correlates with a cumulative weight percentage of bed material less than 0.84 mm (McNeil and Ahnell, 1964) and a weight percentage between 0.125 and 1.0 mm (Murakami et al., 2001). The difference in speciﬁc grain size affecting the permeability coefﬁcient within these studies

Ecological responses to anthropogenic alterations of gravel-bed rivers

515

suggests that permeability may depend on the distribution of grain size but not on a speciﬁc grain size. Embryos and alevins of salmonids remain in porous bed sediment until their emergence to the surface ﬂow, and therefore they are affected by ﬁne sediment accumulation on the streambed. Accordingly, many researchers have investigated the relationship between embryo survival and ﬁne sediment accumulation (e.g., Wickett, 1954; Silver et al., 1963; McNeil and Ahnell, 1964; Wells and McNeil, 1970; Tappel and Bjornn, 1983; Irving and Bjornn, 1984). The successful incubation of salmonid embryos and the emergence of fry thus depend on the extra- and intragravel conditions of spawning grounds, characterized by chemical, physical and hydraulic variables (Chapman, 1988; Bjornn and Reiser, 1991). Chapman (1988) noted in his review that an embryo could be mortally affected by multiple causes. For example, the accumulation of ﬁne sediment results in the depletion of the intragravel dissolved oxygen (DO) concentration because ﬁne sediment lowers the permeability and prevents the interchange between surface and subsurface water. The decrease in intragravel ﬂow rates may directly impair the supply of oxygen and the circulation of water to the redd. Previous studies have reported that the survival and growth rates of embryos decrease with increases of cumulative weight percentage of ﬁne sediment. The former was variously shown to be 6.4 mm (Bjornn, 1968), 4.6 mm (Platts et al., 1979), 2 mm (Hausle and Coble, 1976), 0.85 mm (Cederholm et al., 1981) and 0.84 mm (McNeil and Ahnell, 1964). Field studies by Cederholm et al. (1981) reported that the survival rate of O. kisutch (coho salmon) embryo decreased as much as 10–45%, when the cumulative weight percentage of sediment ﬁner than 0.85 mm exceeded 20% of the bed material. In a case study in Japanese gravel-bed rivers, the survival rate of O. masou (masu salmon) embryo decreased to 20% when the cumulative weight percentage of 2.0 mm (ﬁner than 2 mm) and 1.0 mm (ﬁner than 1 mm) exceeded 40% and 15%, respectively (Yamada, 2002). According to previous studies that examined hydraulic parameters with respect to ﬁne sediment accumulation, embryo survival rates decline with decreasing permeability and apparent velocity in the redd of Oncorhynchus gorbuscha (pink salmon) (Wickett, 1958), Oncorhynchus tshawytscha (chinook salmon) (Gangmark and Bakkala, 1960), O. kisutch and Oncorhynchus mykiss (steelhead trout) (Coble, 1961) and Oncorhynchus nerka (sockeye salmon) (Cooper, 1965). Fine sediment accumulation in the riverbed increases embryo mortality by suffocation, because reduced intragravel ﬂow and permeability limit the amount of DO that would have been supplied to the embryos. Given the close association between DO concentration in the riverbed and surface water, we anticipate that suffocation might become more serious under conditions combining a low DO concentration in surface water and a large amount of ﬁne sediment in the riverbed.

4.2.

Effects of the channel works on openwork gravel bed and benthic fish

Fine sediment runoff has been monitored in many Japanese gravel-bed rivers (e.g., Nakamura et al., 2004). For example, within a reach of the Makomanai River, a

F. Nakamura et al.

516

history of ﬂood- and sediment-related disasters prompted the construction of channel works consisting of 17 small check dams and 21 bed-stabilization structures at intervals of 50 m. These were completed in 1986. A quarry has been operating in the headwaters of this river, and this is considered to be the source of ﬁne sediment. Periphyton growths downstream from the quarry are covered with ﬁne sediment. This becomes obvious in the reach containing the channel works, suggesting the accumulation of ﬁne sediments on stone surfaces (Yamada and Nakamura, 2002) and increasing concern that interstitial habitat in the streambed may be degraded and may be impacting on the benthic ﬁsh (Watanabe et al., 2001). This circumstance provided an opportunity to examine the effects of the channel works on benthic ﬁsh. Three contiguous study reaches were selected and named as the channel works (CW), natural (NA) and bank protection (BP) reaches, in order

C. nozawae density (N/100m2)

200

(a)

r=0.803 (**) rp=0.712 (**) 150

100

50

0 0

10

20

30 Boulder (%)

40

50

60

C. nozawae density (N/100m2)

200 r=0.791(*) rp=0.624 (*)

(b)

150

100

50

0 0

10

20

30 40 50 Openwork gravel (%)

60

70

80

Figure 19.5. Cottus nozawae density in relation to boulder (a) and openwork gravel (b) percentages in Makomanai River, Japan. Symbols ’, m and K denote rifﬂes of channel work (CW), natural (NA) and bank-protection (BP) reach, respectively. r: correlation coefﬁcients, rp: partial correlation coefﬁcients. **Correlation is signiﬁcant at the 0.01 level (two-tailed). *Correlation is signiﬁcant at the 0.05 level (twotailed). (From Watanabe et al., 2001, permission for reproduction obtained from Ecology and Civil Engineering Society, Japan.)

Ecological responses to anthropogenic alterations of gravel-bed rivers

517

from upstream to downstream. In the CW reach, most rifﬂes are glides, and all pools are depressions created by the check dams and other engineering structures. The channel morphology of this reach is uniform. The wetted width of CW is 15 m, which is greater than those for NA and BP, which are less than 10 m. Because of the bed-stabilization structures, the CW reach had a gentle bed slope of 1/100. The area ratio of pool to rifﬂe was as low as 1/7, compared with 1/1.5 to 1/1.6 in the other reaches. In addition, the 60th percentile of the particle-size distribution in the rifﬂe of CW was less than half the size of that in the others. The proportion of openwork gravel was lowest in CW. The density of C. nozawae in the CW reach was signiﬁcantly lower than that in the NA and BP reaches. Figs. 19.5a and b show the relationships between the population density of C. nozawae and the proportion of boulder-size stones and openwork gravel. The density of this species increased with a greater percentage of boulder and openwork gravel. Furthermore, there was a negative relationship between the ﬁne sediment and openwork gravel percentages. These results show the detrimental effect of ﬁne sediment on C. nozawae, suggesting that interstitial spaces in the gravel bed provide a critical habitat for the species. In CW, the tractive force is reduced by check dams and bed-stabilization works. Although this ﬂow condition promotes the deposition of pebble-size stones and stabilizes the channel bed during high ﬂow, it also traps ﬁne sediment during low ﬂow. These physical changes associated with the channel works greatly altered the interstitial habitat required for benthic ﬁsh in the gravel-bed river.

5.

Conclusion

We examined the role of geomorphic disturbance in the creation of various in-stream and ﬂoodplain habitats and in determining their distribution and quality in gravelbed rivers in Japan and East Asia, and described the adaptation of plants and animals to these disturbance-prone environments by their use of various life history strategies. Although these insights have been developed in the past studies, there are very few synthetic reviews of river and ﬂoodplain studies linking between ﬂood disturbance regime and life history traits of certain species, and effects of anthropogenic alterations. The important ecological and geomorphic features in gravel-bed rivers in the three different scales are the shifting habitat mosaic that assures a diversity of habitats at the segment scale, the lateral habitat variation across the river-ﬂoodplain and within the river system at the reach scale, and the role of interstitial spaces in gravel maintained by the bed material mobility during ﬂoods. Most human activity, such as channelization or the construction of reservoirs and dams, greatly reduces the frequency and magnitude of ﬂood disturbances and interrupts the connectivity of the river-ﬂoodplain system. As a result, essential habitats required by aquatic biota at various life stages have been eliminated from the riverine landscape. Flood events are so integral to river-ﬂoodplain ecosystems that the absence of such disturbances can have a fatal effect on aquatic organisms through the loss of their critical habitats and the resulting inadequacy of their life history strategies to adapt.

518

F. Nakamura et al.

References Abe, T., Nakamura, F., 1999. Effects of experimental removal of woody debris on channel morphology and ﬁsh habitats. Ecol. Civ. Eng. 2, 179–190 (in Japanese, with English Abstract). Allan, J.D., Erickson, D.L., Fay, J., 1997. The inﬂuence of catchment land use on stream integrity across multiple spatial scales. Freshw. Biol. 37, 149–161. Angermeier, P.L., Karr, J.R., 1984. Relationships between woody debris and ﬁsh habitat in a small warmwater stream. Trans. Am. Fish. Soc. 113, 716–726. Badri, A.J., Giudicelli, J., Pre´vot, G., 1987. Effects of a ﬂood on the benthic invertebrate community in a Mediterranean river, the Rdat (Morocco). Acta Oecol. 8, 481–500. Barton, B.A., 1977. Short-term effects of highway construction on the limnology of a small stream in southern Ontario. Freshw. Biol. 7, 99–108. Battle, J.M., Mihuc, T.B., 2000. Decomposition dynamics of aquatic macrophytes in the lower Atchafalaya, a large ﬂoodplain river. Hydrobiologia 418, 123–136. Benke, A.C., Van Arsdall, T.C., Gillespie, D.M., Parrish, F.K., 1984. Invertebrate productivity in a subtropical blackwater river: the importance of habitat and life history. Ecol. Monogr. 54, 25–63. Benke, A.C., Wallace, J.B., 2003. Inﬂuence of wood on invertebrate communities in streams and rivers. In: Gregory, S.V., Boyer, K.L., and Gurnell, A.M. (Eds), The Ecology and Management of Wood in World Rivers. American Fisheries Society, Maryland, pp. 149–177. Beschta, R.L., Jackson, W.L., 1979. The intrusion of ﬁne sediments into a stable gravel bed. J. Fish. Res. Board Can. 36, 204–210. Bjornn, T.C., 1968. Survival and emergence of trout and salmon fry in various gravel-sand mixtures, Proceedings of a Forum, Logging and Salmon, American Institute of Fishery Research Biologists, Alaska, pp. 80–88. Bjornn, T.C., Reiser, D.W., 1991. Habitat requirements of salmonids in streams. In: Meehan, W.R. (Ed.), Inﬂuence of Forest and Rangeland Management on Salmonid Fishes and Their Habitats. American Fisheries Society Special Publication 19, American Fisheries Society, Bethesda, Maryland, pp. 83–138. Borchardt, D., 1993. Effects of ﬂow and refugia on drift loss of benthic macroinvertebrates: implications for habitat restoration in lowland streams. Freshw. Biol. 29, 221–227. Bormann, F.H., Likens, G.E., 1979. Pattern and Process in a Forested Ecosystem. Springer-Verlag, Berlin and New York. Brewin, P.A., Newman, T.M.L., Ormerod, S.J., 1995. Patterns of macroinvertebrate distribution in relation to altitude, habitat structure and land use in streams of the Nepalese Himalaya. Arch. Hydrobiol. 135, 79–100. Brookes, A., 1992. River channel change. In: Calow, P. and Petts, G.E. (Eds), The Rivers Handbook. Blackwell, London, pp. 55–75. Brown, A.V., Brussock, P.P., 1991. Comparisons of benthic invertebrates between rifﬂes and pools. Hydrobiologia 220, 99–108. Buffagni, A., Crosa, G.A., Harper, D.M., Kemp, J.L., 2000. Using macroinvertebrate species assemblages to identify river channel habitat units: an application of the functional habitats concept to a large, unpolluted Italian river (River Ticino, northern Italy). Hydrobiologia 435, 213–225. Cederholm, C.J., Reid, L.M. and Salo, E.O. (1981). Cumulative effects of logging road sediment on salmonid populations in Clearwater River, Jefferson Country, Washington, Proceedings of Conference on Salmon Spawning Gravel: A Renewable Resource in the Paciﬁc Northwest, Water Research Center Report 39, Washington State University, Pullman, pp. 38–74. Chapman, D.W., 1988. Critical review of variables used to deﬁne effect of ﬁnes in redds of large salmonids. Trans. Am. Fish. Soc. 117, 1–21. Cline, L.D., Short, R.A., Ward, J.V., 1982. The inﬂuence of highway construction on the macroinvertebrates and epilithic algae of a high mountain stream. Hydrobiologia 96, 149–159. Coble, D.W., 1961. Inﬂuence of water exchange and dissolved oxygen in redd on survival of steelhead trout embryos. Trans. Am. Fish. Soc. 90, 469–474. Cooper, A.C., 1965. The effect of transported stream sediments on survival of sockeye and pink salmon eggs and alevin. International Paciﬁc Salmon Fisheries Commission Bulletin 18. New Westminster, BC.

Ecological responses to anthropogenic alterations of gravel-bed rivers

519

Crisp, D.T., Carling, P.A., 1989. Observations on silting, dimensions and structure of salmonid redds. J. Fish Biol. 34, 119–134. Davies-Colley, R.J., Hickey, C.W., Quinn, J.M., Ryan, P.A., 1992. Effect of clay discharges on streams. 1. Optical properties and epilithon. Hydrobiologia 248, 215–234. Dollar, E.S.J., 2000. Fluvial geomorphology. Prog. Physical Geog. 24, 385–406. Dolloff, C.A., 1986. Effects of stream cleaning on juvenile coho salmon and Dolly Varden in southeastern Alaska. Trans. Am. Fish. Soc. 115, 743–755. Downs, P.W., Thorne, C.R., 1998. Design principles and suitability testing for rehabilitation in a ﬂood defence channel: The River Idle, Nottinghamshire, UK. Aquat. Conserv. Mar. Freshw. Ecosyst. 8, 17–38. Edwards, C.J., Griswold, B.L., Tubb, R.A., et al., 1984. Mitigating effects of artiﬁcial rifﬂes and pools on the fauna of a channelized warmwater stream. North Am. J. Fish. Manage. 4, 194–203. Elliot, S.T., 1986. Reduction of a Dolly Varden population and macrobenthos after removal of logging debris. Trans. Am. Fish. Soc. 115, 392–400. Elser, A.A., 1968. Fish populations of a trout stream in relation to major habitat zones and channel alterations. Trans. Am. Fish. Soc. 97, 389–397. Extence, C.A., 1978. The effects of motorway construction on an urban stream. Environ. Pollut. 17, 245–252. Gangmark, H.A., Bakkala, R.G., 1960. A comparative study of unstable and stable (artiﬁcial channel) spawning streams for incubating king salmon at Mill Creek. Calif. Fish Game 46, 151–164. Gibbs, K.E., Mingo, T.M., 1986. The life history, nymphal growth rates, and feeding habits of Siphlonisca aerodromia Needham (Ephemeroptera: Siphlonuridae) in Maine. Can. J. Zool. 64, 427–430. Gibbs, K.E., Siebenmann, M., 1996. Life history attributes of the rare mayﬂy Siphlonisca aerodromia Needham (Ephemeroptera: Siphlonuridae). J. N. Am. Benthol. Soc. 15, 95–105. Gilvear, D.J., Cecil, J., Parsons, H., 2000. Channel change and vegetation diversity on a low-angle alluvial fan, River Feshie, Scotland. Aquat. Conserv. Mar. Freshw. Ecosyst. 10, 53–71. Gladden, J.E., Smock, L.A., 1990. Macroinvertebrate distribution and production on the ﬂoodplains of two lowland headwater streams. Freshw. Biol. 24, 533–545. Goto, A., 1993. Male mating success and female mate choice in the river sculpin, Cottus nozawae (Cottidae). Environ. Biol. Fish. 37, 347–353. Goto, A., 1994. Life history variations in the sculpin, Cottus nozawae (Cottidae), along the course of small mountain stream. Environ. Biol. Fish. 52, 203–212. Halyk, L.C., Balon, E.K., 1983. Structure and ecological production of the ﬁsh taxocene of a small ﬂoodplain system. Can. J. Zool. 61, 2446–2464. Harris, R.R., Fox, C.A., Risser, R., 1987. Impacts of hydroelectric development on riparian vegetation in the Sierra Nevada region, California, USA. Environ. Manage. 11, 519–527. Harvey, B.C., 1987. Susceptibility of young-of-the-year ﬁshes to downstream displacement by ﬂooding. Trans. Am. Fish. Soc. 116, 851–855. Hausle, D.A., Coble, D.W., 1976. Inﬂuence of sand in redds on survival and emergence of brook trout (Salvelinus fontinalis). Trans. Am. Fish. Soc. 105, 57–63. Hoffmann, A., Hering, D., 2000. Wood-associated macroinvertebrate fauna in central European streams. Int. Rev. Hydrobiol. 85, 25–48. Hohausova´, E., Copp, G.H., Jankovsky´, P., 2001. Movement of ﬁsh between a river and its backwater: diel activity and relation to environmental gradients. Ecol. Freshw. Fish 12, 107–117. Houde, E.D., 1969. Sustained swimming ability of larvae of walleye (Stizostedion vitreum vitreum) and yellow perch (Perca flavescens). J. Fish. Res. Board Can. 26, 1647–1659. Hupp, C.R., Osterkamp, W.R., 1985. Bottomland vegetation distribution along Passage Creek, Virginia, in relation to ﬂuvial landforms. Ecology 66, 670–681. Huryn, A.D., Gibbs, K.E., 1999. Riparian sedge meadows in Maine. In: Batzer, D.P., Rader, R.B., and Wissinger, S.A. (Eds), Invertebrates in Freshwater Wetlands of North America. Wiley, New York, pp. 363–382. Inoue, M., Nakano, S., 1994. Physical environment structure of a small stream with special reference to ﬁsh microhabitat. Jpn. J. Ecol. 44, 151–160 (in Japanese, with English Abstract). Inoue, M., Nakano, S., 1998. Effects of woody debris on the habitat of juvenile masu salmon (Oncorhynchus masou) in northern Japanese streams. Freshw. Biol. 40, 1–16.

520

F. Nakamura et al.

Inoue, M., Nakano, S., 1999. Habitat structure along channel-unit sequences for juvenile salmon: a subunit-based analysis of in-stream landscapes. Freshw. Biol. 42, 597–608. Inoue, M., Nakano, S., Nakamura, F., 1997. Juvenile masu salmon (Oncorhynchus masou) abundance and stream habitat relationships in northern Japan. Can. J. Fish. Aquat. Sci. 54, 1331–1341. Irving, J.S. and Bjornn, T.C., 1984. Effects of substrate size composition on survival of kokanee salmon and cutthroat and rainbow trout embryos. Idaho Cooperative Fishery Research Unit. University of Idaho, Moscow, Technical Report 84-6. Ishikawa, S., 1994. Seedling growth traits of three salicaceous species under different conditions of soil and water level. Ecol. Rev. 23, 1–6. Johnson, W.C., 1994. Woodland expansion in the Platte River, Nebraska: patterns and causes. Ecol. Monogr. 64, 45–84. Johnson, W.C., Burgess, R.L., Keammerer, W.R., 1976. Forest overstory vegetation and environment on the Missouri River ﬂoodplain in North Dakota. Ecol. Monogr. 46, 59–84. Junk, W.J., Bayley, P.B., Sparks, R.E., 1989. The ﬂood pulse concept in river-ﬂoodplain systems. Can. J. Spec. Publ. Fish. Aquat. Sci. 106, 110–127. Karrenberg, S.P., Edwards, P.J., Kollman, J., 2002. The life history of Salicaceae living in the active zone of ﬂoodplains. Freshw. Biol. 47, 733–748. Kawaguchi, Y., Nakamura, F., Kayaba, Y., 2005. Effects of a re-meandering project on the physical habitats and ﬁsh in the Shibetsu River. Ecol. Civil Eng. 7, 187–199 (in Japanese, with English Abstract). Kawaguchi, Y., Saiki, M., Mizuno, T. and Kayaba, Y., 2006. Effects of different bank types on aquatic organisms in an experimental stream: contrasting vegetation with concrete revetment. Verhandlungen der Internationalen Vereinigung fur Theoretische und Angewandte Limnologie. 29, 1427–1432. Kon, H., Okitsu, S., 1999. Role of land-surface disturbance in regeneration of Ulmus davidiana var. japonica in a cool temperate deciduous forest on Mt. Asama, central Japan. J. Jap. Forest Soc. 81, 29–35 (in Japanese, with English summary). Lancaster, J., Hildrew, A.G., 1993. Flow refugia and the microdistribution of lotic macroinvertebrates. J. N. Am. Benthol. Soc. 12, 385–393. Lee, S., Fujita, K., Tsukahara, T., et al., 1998. Roles of ﬂoodsand ﬁne sediment transport to woodland expansion on a gravel river-bed. Annu. J. Hydraul. Eng., JSCE 42, 433–438 (in Japanese, with English Abstract). Lehmkuhl, D.M., Anderson, N.H., 1972. Microdistribution and density as factors affecting the downstream drift of mayﬂies. Ecology 53, 661–667. Marchese, M., Ezcurra de Drago, I., 1992. Benthos of the lotic environments in the middle Parana´ River system: transverse zonation. Hydrobiologia 237, 1–13. Matthaei, C.D., Arbuckle, C.J., Townsend, C.R., 2000. Stable surface stones as refugia for invertebrate during disturbance in a New Zealand stream. J. N. Am. Benthol. Soc. 19, 82–93. Matthaei, C.D., Townsend, C.R., 2000. Inundated ﬂoodplain gravels in a stream with an unstable bed: temporary shelter or true invertebrate refugium? N. Z. J. Mar. Freshw. Res. 34, 147–156. McNeil, W.J. and Ahnell, W.H., 1964. Success of pink salmon spawning relative t size of spawning bed materials. U.S. Fish and Wildlife Service Spatial Scientiﬁc Report Fisheries 469. Michener, W.K., Haeuber, R.A., 1998. Flooding: natural and managed disturbances. BioScience 48, 677–680. Modde, T., Ford, R.C., Parsons, M.G., 1991. Use of habitat-based stream classiﬁcation system for categorizing trout biomass. North Am. J. Fish. Manage. 11, 305–311. Moore, K.M.S., Gregory, S.V., 1988. Summer habitat utilization and ecology of cutthroat trout fry (Salmo clarki) in Cascade mountain streams. Can. J. Fish. Aquat. Sci. 45, 1921–1930. Morisawa, M., 1985. Rivers, Form and Process, Geomorphology Texts 7. Longman, London. Murakami, M., Yamada, H., Nakamura, F., 2001. Hydraulic conductivity of substrate and openwork gravel rate associated with ﬁne sediment deposition in mountain streams, southern Hokkaido. Ecol. Civil Eng. 4, 109–120 (in Japanese, with English Abstract). Nagasaka, A., Nakamura, F., 1999. The inﬂuence of land-use changes on hydrology and riparian environment in a northern Japanese landscape. Landsc. Ecol. 14, 543–556. Nagasaka, Y., 1996. Salicaceae species inhabit at riparian zone. Koshunai-Kihou 101, 12–17 (in Japanese). Nakamura, F., 1990. Perspectives for the effects of geomorphological processes on forest ecosystems. Biol. Sci. (Tokyo) 42, 57–67 (in Japanese).

Ecological responses to anthropogenic alterations of gravel-bed rivers

521

Nakamura, F., Kameyama, S., Mizugaki, S., 2004. Rapid shrinkage of Kushiro Mire, the largest mire in Japan, due to increased sedimentation associated with land-use development in the catchment. Catena 55, 213–229. Nakamura, F., Maita, H., Araya, T., 1995. Sediment routing analyses based on chronological changes in hillslope and riverbed morphologies. Earth Surf. Process. Landf. 20, 333–346. Nakamura, F., Shin, N., 2001. The downstream effects of dams on the regeneration of riparian tree species in northern Japan. American Geophysical Union monograph on ‘‘Geomorphic Processes and Riverine Habitat’’. Water Sci. Appl. 4, 173–181. Nakamura, F., Yajima, T., Kikuchi, S., 1997. Structure and composition of riparian forests with special reference to geomorphic site conditions in the Tokachi River, northern Japan. Plant Ecol. 133, 209–219. Nakano, D. and Nakamura, F., 2006. Lateral variation of lotic macroinvertebrate community from the headwater to lowland meandering stream in the Nishibetsu River basin, northern Japan. Verhandlungen der Internationalen Vereinigung fur Theoretische und Angewandte Limnologie. 29,1377–1382. Nakano, D., Nunokawa, M., Nakamura, F., 2005. Changes in distribution and structure of macroinvertebrate community before and after re-meandering experiment. Ecol. Civil Eng. 7, 173–186 (in Japanese, with English Abstract). Nanson, G.C., Beach, H.F., 1977. Forest succession and sedimentation on a meandering-river ﬂoodplain, northeast British Columbia, Canada. J. Biogeogr. 4, 229–251. Negishi, J.N., Inoue, M., Nunokawa, M., 2002. Effects of channelisation on stream habitat in relation to a spate and ﬂow refugia for macroinvertebrates in northern Japan. Freshw. Biol. 47, 1515–1529. Newcombe, C.P., MacDonald, D.D., 1991. Effects of suspended sediments on aquatic ecosystems. North Am. J. Fish. Manage. 11, 72–82. Niiyama, K., 1987. Distribution of Salicaceae species and soil texture of habitats along the Ishikari River. Jpn. J. Ecol. 37, 163–174 (in Japanese, with English Abstract). Niiyama, K., 1989. Distribution of Chosenia arbutifolia and soil texture of habitats along the Satsunai River. Jpn. J. Ecol. 39, 173–182 (in Japanese, with English Abstract). Niiyama, K., 1990. The role of seed dispersal and seedling traits in colonization and coexistence of Salix spp. in a seasonally ﬂooded habitat. Ecol. Res. 5, 317–332. Niiyama, K., 1995. Life history traits of salicaceous species and riparian environment. Jpn. J. Ecol. 45, 301–306 (in Japanese). Niiyama, K., 2002. Riparian forests (Tentative title translated from Japanese to English by the present authors). In: Sakio, H. and Yamamoto, F. (Eds), Ecology of Riparian Forests. University of Tokyo Press, Tokyo, Japan, pp. 61–93(in Japanese). Nilsen, H.C., Larimore, R.W., 1973. Establishment of invertebrate communities on log substrates in the Kaskaskia River, Illinois. Ecology 54, 366–374. Ormerod, S.J., Rundle, S.D., Lloyd, E.C., Douglas, A.A., 1993. The inﬂuence of riparian management on the habitat structure and macroinvertebrate communities of upland streams draining plantation forests. J. Appl. Ecol. 30, 13–24. Osterkamp, W.R., Hupp, C.R., 1984. Geomorphic and vegetative characteristics along three northern Virginia streams. Geol. Soc. Am. Bull. 95, 1093–1101. Palmer, M.A., Arensburger, P., Martin, A.P., Denman, D.W., 1996. Disturbance and patch-speciﬁc responses: the interactive effects of woody debris and ﬂoods on lotic invertebrates. Oecologia 105, 247–257. Penczak, T., Kruk, A., Koszalin´ski, H., et al., 2000. Fishes of three oxbow lakes and their parent Pilica River: 25 years later. Pol. Arch. Hydrobiol. 47, 115–130. Penczak, T., Zieba, G., Koszalin´ski, H., Kruk, A., 2003. The importance of oxbow lakes for ﬁsh recruitment in a river system. Arch. Hydrobiol. 158, 267–281. Pie´gay H., Grant G., Nakamura F., Trustrum N., 2006. Braided river management: from assessment of river behavior to improved sustainable development. In: Sambrook-Smith, G.H., Best, J.L., Bristow, C.S., and Petts, G.E. (Eds), Braided Rivers: Process, Deposits, Ecology and Management. Special publication 36 of the International Association of Sedimentologists. Platts, W.S., Shirazi, M.A., Lewis, D.H., 1979. Sediment particle sizes used by salmon for spawning with methods for evaluation. U.S. Environmental Protection Agency, Corvallis, OR, (EPA 600/3-79-043).

522

F. Nakamura et al.

Platts, W.S., Torquemada, R.J., McHenry, M.L., Graham, C.K., 1989. Changes in salmon spawning and rearing habitat from increased delivery of ﬁne sediment to the South Fork Salmon River, Idaho. Trans. Am. Fish. Soc. 118, 274–283. Poff, N.L., Ward, J.V., 1989. Implications of streamﬂow variability and predictability for lotic community structure: a regional analysis of streamﬂow patterns. Can. J. Fish. Aquat. Sci. 46, 1805–1818. Reily, P.W., Johnson, W.C., 1982. The effects of altered hydrologic regime on tree growth along the Missouri River in North Dakota. Can. J. Bot. 60, 2410–2423. Rempel, L.L., Richardson, J.S., Healey, M.C., 1999. Flow refugia for benthic macroinvertebrates during ﬂooding of a large river. J. N. Am. Benthol. Soc. 18, 34–48. Rempel, L.L., Richardson, J.S., Healey, M.C., 2000. Macroinvertebrate community structure along gradients of hydraulic and sedimentary conditions in a large gravel-bed river. Freshw. Biol. 45, 57–73. Resh, V.H., Brown, A.V., Covich, A.P., et al., 1988. The role of disturbance in stream ecology. J. N. Am. Benthol. Soc. 7, 433–455. Richards, C., Host, G.E., Arthur, J.W., 1993. Identiﬁcation of predominant environmental factors structuring stream microinvertebrate communities within a large agricultural catchment. Freshw. Biol. 29, 285–294. Robinson, C.T., Tockner, K., Ward, J.V., 2002. The fauna of dynamic riverine landscapes. Freshw. Biol. 47, 661–677. Robinson, C.T., Uehlinger, U., Monaghan, M.T., 2003. Effects of a multi-year experimental ﬂood regime on macroinvertebrates downstream of a reservoir. Aquat. Sci. 65, 210–222. Robinson, C.T., Uehlinger, U., Monaghan, M.T., 2004. Stream ecosystem response to multiple experimental ﬂoods from a reservoir. River Res. Appl. 20, 359–377. Rood, S.B., Heinze-Milne, S., 1989. Abrupt downstream forest decline following river damming in southern Alberta. Can. J. Bot. 67, 1744–1749. Sagawa, S., Kayaba, Y., Arai, H., Amano, K., 2005. Habitat selection by larval cyprinid ﬁshes: Relationship between larval habitats and experimental ﬂooding. Ecol. Civil Eng. 7, 129–138 (in Japanese, with English Abstract). Sato, H., 1995. Studies on the dynamics of Pterocarya rhoifolia forest in southern Hokkaido. Bull. Hokkaido For. Res. Inst. 32, 55–96 (in Japanese, with English Abstract). Scha¨lchli, U., 1992. The clogging of coarse gravel river beds by ﬁne sediment. Hydrobiologia 235/236, 189–197. Scott, M.T., Nielsen, L.A., 1989. Young ﬁsh distribution in backwaters and main-channel borders of the Kanawha River, West Virginia. J. Fish Biol. 35, 21–27. Seiwa, K., 1997. Variable regeneration behaviour of Ulmus davidiana var. japonica in response to disturbance regime for risk spreading. Seed Sci. Res. 7, 195–207. Shimatani, Y., Oguri, Y., Kayaba, Y., 1994. Impact of stream modiﬁcation on habitat component and ﬁsh in Tagawa River. Annu. J. Hydraul. Eng., JSCE 38, 337–344 (in Japanese, with English Abstract). Shin, N., Ishikawa, S., Iwata, S., 1999. The mosaic structure of riparian forest and its formation pattern along the Azusa River, Kamikochi, Central Japan. Jpn. J. Ecol. 49, 71–81 (in Japanese, with English Abstract). Shin, N., Nakamura, F., 2005. Effects of ﬂuvial geomorphology on riparian tree species in Rekifune River, northern Japan. Plant Ecol. 178, 15–28. Shirvell, C.S., 1990. Role of instream rootwads as juvenile Coho salmon (Oncorhynchus kisutch) and Steelhead trout (O. mykiss) cover habitat under varying streamﬂows. Can. J. Fish. Aquat. Sci. 47, 852–861. Silver, S.J., Warren, C.E., Doudoroff, P., 1963. Dissolved oxygen requirements of developing steelhead trout and Chinook salmon embryos at different water velocities. Trans. Am. Fish. Soc. 92, 327–343. Smock, L.A., 1999. Riverine ﬂoodplain forests of the southeastern United States. In: Batzer, D.P., Rader, R.B., and Wissinger, S.A. (Eds), Invertebrates in Freshwater Wetlands of North America. Wiley, New York, pp. 137–165. Stanford, J.A., Ward, J.V., 1993. An ecosystem perspective of alluvial rivers: connectivity and the hyporheic corridor. J. N. Am. Benthol. Soc. 12, 48–60. Takagi, M., Nakamura, F., 2003. The downstream effects of water regulation by the dam on the riparian tree species in the Satsunai River. J. Jap. Forest Soc. 85, 214–221 (in Japanese, with English Abstract).

Ecological responses to anthropogenic alterations of gravel-bed rivers

523

Takahashi, G., Higashi, S., 1984. Effect of channel alteration on ﬁsh habitat. Jpn. J. Limnol. 45, 178–186. Tappel, P.D., Bjornn, T.C., 1983. A new method of relating size of spawning gravel to salmonid embryo survival. North Am. J. Fish. Manage. 3, 123–135. Toyoshima, T., Nakano, S., Inoue, M., et al., 1996. Fish population responses to stream habitat improvement in a concrete-lined channel. Jpn. J. Ecol. 46, 9–20 (in Japanese, with English Abstract). Urabe, H., Nakano, S., 1998. Contribution of woody debris to trout habitat modiﬁcation in small streams in secondary deciduous forest, northern Japan. Ecol. Res. 13, 335–345. Van Nieuwenhuyse, E.E., LaPerriere, J.D., 1986. Effects of placer gold mining on primary production in sub arctic streams of Alaska. Water Resour. Bull. 22, 91–99. Wallace, J.B., Anderson, N.H., 1996. Habitat, life history, and behavioral adaptations of aquatic insects. In: Merritt, R.W. and Cummins, K.W. (Eds), An Introduction to the Aquatic Insects of North America. Kendall/Hunt Publishing Company, Iowa, pp. 41–73. Walling, D.E., 1990. Linking the ﬁeld to the river: sediment delivery from agricultural land. In: Boardman, J., Foster, I.D.L., and Dearing, J.A. (Eds), Soil Erosion on Agricultural Land. John Wiley, Chichester, pp. 129–152. Ward, J.V., Bretschko, G., Brunke, M., et al., 1998. The boundaries of river systems: the metazoan perspective. Freshw. Biol. 40, 531–569. Ward, J.V., Tockner, K., Arscott, D.B., Claret, C., 2002. Riverine landscape diversity. Freshw. Biol. 47, 517–539. Watanabe, K., Nakamura, F., Kamura, K., et al., 2001. Inﬂuence of stream alteration on the abundance and distribution of benthic ﬁsh. Ecol. Civil Eng. 4, 133–146 (in Japanese, with English Abstract). Waters, T.F., 1995. Sediment in Streams – Sources, Biological Effects, and Control. American Fisheries Society Monograph 7, American Fisheries Society, Bethesda, Maryland. Wells, R.A. and McNeil, W.J., 1970. Effect of quality of the spawning bed on the growth and development of pink salmon embryos and alevins. U.S. Fish and Wildlife Service Spatial Scientiﬁc Report Fisheries 616. Wetzel, R.G., 2001. Limnology. Academic Press, London. Wickett, W.P., 1954. The oxygen supply to salmon eggs in spawning beds. J. Fish. Res. Board Can. 11, 933–953. Wickett, W.P., 1958. Review of certain environmental factors affection the production of pink and chum salmon. J. Fish. Res. Board Can. 15, 1103–1126. Wilcock, D.N., Essery, C.I., 1991. Environmental impacts of channelization on the River Main, County Antrim, Northern-Ireland. J. Environ. Manage. 32, 127–143. Winemiller, K.O., Tarim, S., Shormann, D., Cotner, J.B., 2000. Fish assemblage structure in relation to environmental variation among Brazos River oxbow lakes. Trans. Am. Fish. Soc. 129, 451–468. Wondzell, S.M., Swanson, F.J., 1999. Floods, channel change, and the hyporheic zone. Water Resour. Res. 35, 555–567. Wood, P.J., Armitage, P.D., 1997. Biological effect of ﬁne sediment in the lotic environment. Environ. Manage. 21, 203–217. Yamada, H., 2002. Effects of ﬁne sediment deposition on the lotic biota associated with changing mechanisms of riverbed structure. Doctoral dissertation, Hokkaido University, p. 136 (in Japanese). Yamada, H., Nakamura, F., 2002. Effect of ﬁne sediment deposition and channel works on periphyton biomass in the Makomanai River, northern Japan. River Res. Appl. 18, 481–493. Yamada, H., Nakamura, F., Watanabe, Y., et al., 2005. Measuring hydraulic permeability in a streambed using the packer test. Hydrol. Process. 19, 2507–2524.

Gravel-Bed Rivers VI: From Process Understanding to River Restoration H. Habersack, H. Pie´gay, M. Rinaldi, Editors r 2008 Elsevier B.V. All rights reserved.

525

20 A review on channel incision in the Polish Carpathian rivers during the 20th century Bart"omiej Wyz˙ ga

Abstract Rivers draining the Polish Carpathians deeply incised over the 20th century and in many sections, the downcutting was especially rapid in the second half of the century. Incision has resulted from the increase in transport capacity of the rivers caused by their channelization, and the concomitant decrease in sediment supply to the channels. In some of the rivers, in-stream gravel mining has additionally reduced the amount of sediment available for ﬂuvial transport. Where the rivers had insufﬁcient energy to destroy the river-control structures and remained laterally stable following their channelization, bed degradation has proceeded at a relatively steady rate. On the high-energy rivers, the periods of incision of the regulated channel alternated with the periods of lateral channel migration following the destruction of channelization structures. The main phase of incision of the Carpathian rivers occurred progressively later in the upstream direction, this reﬂecting the variation in timing of the most intense channelization works along their course, the operation of upstreamprogressing bed degradation as well as the concentration of the land use changes from the second half of the century in the montane parts of the catchments. A marked increase in ﬂood hazard to downstream reaches and a reduction in the potential of Carpathian ﬂoodplains for sediment storage have been the most important detrimental effects of the river incision manifested at the regional scale. Changes in management of the rivers are necessary to reduce their transport capacity and re-establish the conditions for water and sediment storage on the ﬂoodplains. 1.

Introduction

In alluvial rivers, temporal trends in vertical channel position reﬂect the mutual relationship between transport capacity of the river and the availability of bed material for ﬂuvial transport. During the 19th century rivers draining the Polish Carpathians showed a marked tendency to aggrade (Wyz˙ ga, 1993a). The aggradation of E-mail address: [email protected] ISSN: 0928-2025

DOI: 10.1016/S0928-2025(07)11142-1

526

B. Wyz˙ga

the river beds was accompanied by widening of the channels and, at many locations, by the development of multi-thread channel pattern (Fig. 20.1) (Klimek and Trafas, 1972; Szuman´ski, 1986; Wyz˙ ga, 1993a). Channel bar deposits from that time consisted of overloose and normally loose gravels (cf. Church, 1978) and lacked armour layers. The aggrading tendency of the channels during the 19th century testiﬁes to an overloading of the Carpathian rivers with sediment, whereas the character of the channel sediments is indicative of high rates of bedload transport and deposition of the materials by ﬂood waves with high peak discharges and short time bases (Wyz˙ ga, 1993a, 2001a). In the 20th century, the Carpathian rivers showed the opposite tendency (Fig. 20.1) (Punzet, 1981; Klimek, 1983; Wyz˙ ga, 1991, 2001a). Bed degradation and the resultant channel deepening have signiﬁcantly altered morphology and functioning of the watercourses, and a variety of detrimental effects of incision have been identiﬁed in the channels and on the valley ﬂoors of the Carpathian rivers (Froehlich, 1980; Klimek, 1983; Wyz˙ ga, 1991, 2001a). Degradational channel tendency during the 20th century, especially during the last decades, was documented for many rivers of the world. In some cases, it was possible to ascribe channel incision to a single or dominating, disturbing factor such as: gravel mining (Bull and Scott, 1974; Collins and Dunne, 1989; Sear and Archer, 1998), dams (Williams and Wolman, 1984), river channelization (Emerson, 1971; Brookes, 1987; Simon, 1989), catchment reafforestation (Lie´bault and Pie´gay, 2001) or cessation of the in-channel deposition of mine tailings (Knighton, 1989). However, much more common are the situations where incision has resulted from a combined impact of two or

Figure 20.1. Structure of a cutbank of the middle Raba at Winiary showing a sedimentary record of the aggradational/degradational tendencies of the river channel from the last two centuries. The shallow braid was eroded in the upper part of the sequence of overbank deposits and ﬁlled with massive gravel at about the turn of the 20th century, with the culmination of the aggradational river tendency. A high position of the braid above the low-water level in the contemporary channel testiﬁes to the rapid incision of the Raba over the last century. Marks on the rope stretched along the cutbank are spaced at 1 m intervals.

A review on channel incision in the Polish Carpathian rivers

527

more disturbances, frequently operating at different temporal and spatial scales (e.g., Bravard et al., 1997; Landon et al., 1998; Rinaldi, 2003; Surian and Rinaldi, 2003). This paper utilizes data on vertical channel changes from numerous water-gauge stations operating in the rivers draining the Polish Carpathians to conclude about the dimensions and course of incision of the rivers during the 20th century. It also brings together extensive information from previous research on the phenomenon to infer about its causes and their relative importance. Finally, it presents the effects of incision of the Carpathian rivers, especially those apparent at the regional scale. This regional case study may be interesting for the gravel-bed river scientiﬁc community because recognition of unfavourable effects of rapid channel downcutting, wherever it occurs, is a pre-requisite to formulating appropriate remedial measures and undertaking actions to restore incised rivers, whereas identiﬁcation of the causes of incision is essential to succeed in restoration efforts.

2.

Physical setting and historical background

Most rivers in the Carpathian part of the upper Vistula River drainage basin rise in the Beskidy Mountains which are underlain by ﬂysch and range up to 1725 m a.s.l. in the western part of the region and to 1346 m a.s.l. in its eastern part. Only the Dunajec River and some of its major tributaries originate in the high-mountain Tatra massif with a considerable proportion of crystalline rocks in its structure and elevations ranging up to 2655 m a.s.l. (Fig. 20.2). There are marked differences between the rivers draining the western and eastern part of the Polish Carpathians (Klimek, 1979), which reﬂect distinct physiography of the catchments in both areas (Fig. 20.2). Mountain areas predominate in the western part; here, the rivers have steep channel gradients, ﬂow in cobble to pebble gravels and are characterized by high stream power at ﬂood ﬂows. In the eastern part, the main Carpathian rivers have long reaches within the foothill and foreland areas, where they are typiﬁed by gentler channel gradients, ﬁner calibre of their bed material and lower stream power values at ﬂood ﬂows. Within the upper Vistula River drainage basin, annual precipitation totals range from 1200 to 1900 mm in the Tatra massif to 600–700 mm in the Carpathian foreland (Niedz´wiedz´ and Obr˛ebska-Starklowa, 1991), with the respective values of the coefﬁcient of runoff varying from more than 60% in the Tatran watersheds to 20–30% in the foreland area (Dynowska, 1991). In the western part of the drainage basin, with a relatively high frequency of the arrival of oceanic air-masses, ﬂoods originate mostly due to summer rains. In the eastern part with more continental climate, the Carpathian rivers are characterized by frequent ﬂoods of moderate magnitude following snow melt, and rare, large ﬂoods caused by summer rains. Three economic and demographic factors seem to have exerted a substantial inﬂuence on the 20th-century evolution of the Carpathian rivers. First, at the end of the 19th century, the Carpathian part of the upper Vistula drainage basin was a densely populated, agricultural region (Pietrzak, 2005). This situation created a great demand for reclamation of riparian areas and protection of the valley ﬂoors from ﬂooding that resulted in undertaking intense river-control works early in the 20th century (K˛edzior, 1928). Second, large amounts of aggregate were required with

B. Wyz˙ga

528

Figure 20.2. Dimensions of channel incision of Carpathian tributaries to the Vistula during the 20th century and in its second half inferred from the lowering of minimum annual water stage at gauging stations on the rivers. (1) High mountains; (2) mountains of intermediate and low height; (3) foothills; (4) intramontane and submontane depressions; (5) uplands; (6) lowering of minimum annual stage at water-gauge stations during the 20th century (grey semicircles) and in its second half (black semicircles).

rapid industrialization and urbanization of southern Poland after the World War II. Since the alluvium of the Carpathian rivers was the only available source of gravel in the region, over at least two post-war decades the need was being satisﬁed by the sediments mined from the river channels (Rinaldi et al., 2005). Third, the eastern part of the Polish Carpathians was dramatically depopulated in the mid-1940s and reafforested thereafter. This has signiﬁcantly limited sediment delivery to the river channels in that area (Lach and Wyz˙ ga, 2002).

3.

Study methods

In the foothill and foreland reaches of Carpathian tributaries to the Vistula, a number of water-gauge stations have been operating since the last decades of the 19th century (Fig. 20.2). The amount and timing of 20th-century changes in vertical channel position in the reaches were reconstructed by examining the record of the lowest annual water stages at the stations (Punzet, 1981; Klimek, 1983; Wyz˙ ga, 1991, 2001a). Because the observed variation of the lowest ﬂows of the rivers has been shown to explain some deviations from general trends of the lowest stages only, not the trends themselves (Wyz˙ ga, 1997), multi-year variations of minimum annual water stages can reasonably be considered as reﬂecting erosional or aggradational tendencies of the river channels.

A review on channel incision in the Polish Carpathian rivers

529

In the montane reaches of Carpathian rivers, a small number of gauging stations operated throughout the 20th century (Fig. 20.2) and this makes dating the onset of channel incision and establishing its total extent more difﬁcult than in the foothill and foreland river reaches. Scarce hydrometric data from the montane reaches were supplemented by information obtained from examination of the difference in elevation between the beds of cartographically dated paleochannels and contemporary river beds as well as between the tops of older and contemporary gravel bars (Zawiejska and Wyz˙ ga, 2005). Planar and cross-sectional channel changes of Carpathian rivers during the 20th century were documented in a number of studies (e.g., Krzemien´, 1981; Szuman´ski, 1986; Klimek, 1987; Wyz˙ ga, 1991, 1993a–c, 2001a) which utilized information from historical maps, aerial photos and channel regulation plans as well as from repeated channel cross-section surveys at water-gauge stations performed by the Hydrologic Survey. Facies patterns of channel sediments of the middle Raba River dated on the basis of cartographic data were studied in river cutbanks and walls of gravel pits, and compared with those of contemporary gravel-bar sediments to infer about depositional conditions during the passage of ﬂood waves (Wyz˙ ga, 1993a–c, 2001a). Samples were collected from a short valley reach to avoid distortion of temporal sediment-change patterns by downstream trends. Temporal trends in channel sedimentation similar to those described from the middle Raba were also observed in other rivers of the Polish Carpathians (e.g., Lach and Wyz˙ ga, 2002). The location and timing of large-scale, in-stream gravel mining, the volumes of extracted material as well as the post-mining channel adjustments were documented for two Carpathian rivers in which such activity was especially pronounced (Augustowski, 1968; Osuch, 1968; Rinaldi et al., 2005). The effect of channel incision upon ﬂood ﬂows was investigated using a set of procedures which analyse temporal trends in the relationship between inﬂow and outﬂow peak discharges of ﬂood waves passing the modiﬁed reach, and relate them to alterations in vertical channel position (Wyz˙ ga, 1996, 1997). In the present article, a simple comparison of the magnitude of ﬂoods of given recurrence intervals at two gauging stations on the Wis"oka River is presented for the record periods characterized, respectively, by small and high degree of channel incision between the stations. Finally, the impact of channel incision on the potential of Carpathian ﬂoodplains for sediment storage was studied by analysing temporal changes in ﬂow characteristics at representative gauge cross-sections located on the rivers from the western and eastern part of the Polish Carpathians. The analysed parameters comprised: (i) percentage of the total ﬂow conveyed in the extra-channel zone of the crosssections at particular ﬂood discharges; (ii) frequency of the valley ﬂoor inundation at given discharges; (iii) relative elevation of ﬂood stages above the channel bed; and (iv) ﬂow-velocity in the extra-channel zone (Wyz˙ ga, 2001b; Lach and Wyz˙ ga, 2002).

4.

Amount of the 20th-century incision of the Polish Carpathian rivers

In the middle and lower courses of Carpathian tributaries to the Vistula, the onset of channel downcutting typically occurred around the turn of the 20th century

B. Wyz˙ga

530

(Punzet, 1981; Wyz˙ ga, 1991) and to date, the rivers have incised by 1.3–3.8 m as shown by the lowering of their minimum annual stages at water-gauge stations (Fig. 20.2). Sections with about 3 m of channel downcutting accomplished over the century occur on most of the rivers and the greatest extent of incision is observed at the Łabuzie and Brzez´nica stations on the Wis"oka River. In many sections the downcutting was especially rapid in the second half of the century (Fig. 20.2). During the second half of the century, channel downcutting was also recorded in the upper course of some Carpathian tributaries to the Vistula (Lach and Wyz˙ ga, 2002; Krzemien´, 2003; Kukulak, 2003) and in their mountain tributaries (Soja, 1977; Froehlich, 1982; Rinaldi et al., 2005). Here, at most of the stations for which a stage record from the whole 20th century is available, the main degradational phase took place during the second half of the century. In the extreme case, 2.8 m of the total channel downcutting accomplished at the Z˙ o´"ko´w station on the upper Wis"oka River over the 20th century, about 2.3 m occurred between 1964 and the early 1980s, with the average rate of incision in that period exceeding 10 cm/year (Fig. 20.2, see also Fig. 20.7 below) (Lach and Wyz˙ ga, 2002). The hydrometric data from gauging stations (Fig. 20.2) as well as ﬁeld evidence indicate that 0.5–3.5 m of bed degradation has occurred in the montane river reaches, its intensity having been highly varied spatially. There also occur few unregulated or bedrock-controlled channel sections that remained vertically stable over recent decades (Zawiejska and Wyz˙ ga, 2005). At many locations in the upper course of main Carpathian rivers and along their mountain tributaries, incision has resulted in dissection of the whole thickness of alluvium and transformation of the former alluvial channels into bedrock ones (Lach and Wyz˙ ga, 2002; Krzemien´, 2003; Rinaldi et al., 2005). Where the rivers are underlain by sandstones, bed lowering has ceased following the change to bedrock boundary conditions. However, incision has continued in channel sections underlain by less resistant lithologies and up to 3.5 m of bed lowering was recorded in the section of the Czarny Dunajec River where the bedrock consists of unconsolidated Pliocene clays (Zawiejska and Wyz˙ ga, 2005).

5.

Causes of the channel incision

Rapid incision of the rivers draining the Polish Carpathians indicates that in the 20th century their transport capacity must have greatly exceeded sediment delivery to the channels. A few factors have contributed to this situation, different from that prevalent during the 19th century. 5.1.

Increase in transport capacity of the rivers due to channelization works

Although early localized attempts to channelize main rivers of the Polish Carpathians were undertaken already in the second half of the 19th century, intensive and widespread channelization works on the rivers began in 1904 and were continued through the 1930s (K˛edzior, 1928). In that period, works were concentrated in the middle and lower courses of Carpathian tributaries to the Vistula River. They

A review on channel incision in the Polish Carpathian rivers

531

consisted of channel straightening through meander cut-offs, channel narrowing by groynes and the lining of concave banks by gabions and rip-rap; moreover, channel stretches where the ﬂow diverged among mid-channel bars or islands were replaced by a single, artiﬁcial channel. In the foreland river reaches, the channelization works either immediately succeeded or were concurrent with the construction of ﬂoodprotection embankments. The ﬁrst phase of channelization had the greatest impact on the foreland river reaches. For example, the foreland reach of the Dunajec River was shortened by about 10% and channel width was here reduced by one third (Zawiejska and Wyz˙ ga, 2005). In the foreland reach of the Raba, the channel was shortened by 15% and also considerably narrowed; in the foothill reach, little change in the width and length of the channel took place although the works helped to train the thalweg and led to the formation of a single-thread channel by 1932 (Wyz˙ ga, 1991, 1993a). After a break in channelization works during and immediately after the World War II, the works were resumed late in the 1950s. In the second half of the century, they were concentrated in the middle and upper courses of main rivers of the Polish Carpathians and in their mountain tributaries. In the foothill reaches of the Carpathian tributaries to the Vistula River, the second phase of channelization caused considerable narrowing of the channels and shortened their length. For example, between 1955 and 1987, the Dobczyce-Gdo´w reach of the Raba River was shortened by 15% and the average channel width was here reduced from 140 to 60 m (Wyz˙ ga, 2001a). If the pre-regulation and regulated channels in this reach had the same ﬂow capacity, the changes would have increased unit stream power at bankfull ﬂow by about 170%. In fact, the increase in unit stream power must have been even greater as ﬂow capacity of regulated Carpathian channels was typically increased in comparison with the pre-regulation conditions (with regulated channels designed to convey 2- to 5-year ﬂows in rural areas and 10-year ﬂows in urban areas (Raczyn´ski, 1989)). In the montane reaches of the Carpathian rivers, the channels were narrowed and most multi-thread river sections were replaced by a single, artiﬁcial channel (Krzemien´, 1981; Zawiejska and Krzemien´, 2004; Zawiejska and Wyz˙ ga, 2005). The shortening and narrowing of the channels in the course of channelization have increased unit stream power of the rivers. At the same time, the concentration of ﬂow which was previously divided between separate braids or channels in multi-thread river reaches must have reduced channel-form resistance (cf. Bathurst, 1982), thus increasing a portion of the total ﬂow energy available for sediment transport. The combined effect of the changes has been the increase in mean velocity at a given ﬂow (Wyz˙ ga, 1993b, 2001a) and therefore in transport capacity of the rivers.

5.2.

Reduction in sediment delivery to the rivers

Contemporary channel bar deposits of the rivers draining the Polish Carpathians are coarser-grained and better sorted than their 19th-century counterparts; moreover, contemporary bar surfaces are typically armoured, in contrast to the 19th-century bar sediments lacking armour layers (Wyz˙ ga, 1993a–c, 2001a). However, it might be argued that the altered mode of channel sedimentation reﬂects channelization-induced

532

B. Wyz˙ga

changes in hydraulic conditions in the rivers, rather than changes in sediment supply to their channels. Undoubtedly, the latter must have also contributed to the sedimentary change and their inﬂuence can be demonstrated by comparing channel geometry and sediments of different age that originated under free channel development conditions. Such data were presented (Wyz˙ ga, 2001a) with respect to the middle Raba for the second half of the 19th century, before the initiation of channelization works, and for the mid-20th century, when the river destroyed the regulation structures and formed a natural channel in its middle course during the break in channelization works. In the second half of the 19th century, the middle Raba was a wide and shallow river. In the Dobczyce-Gdo´w reach, it ﬂowed in a straight or braided channel (sinuosity index SI ¼ 1.12). In contrast, the unregulated channel from the mid-20th century showed a conspicuous tendency to meander. Between 1932 and 1955 the river increased its length in the reach by 22% with the resultant increase in sinuosity (to SI ¼ 1.31) and reduction in gradient, and the changes were accompanied by the increase in channel depth compensating for the decrease in river gradient. The average distance between successive points of ﬂow inﬂection in the 1955 channel was reduced by almost half in comparison with that typifying the pre-regulation channel from 1878. The changes in river planform were accompanied by alterations in crosssectional channel geometry; at the Gdo´w gauging station, the width/depth ratio of the Raba channel diminished from 40 in 1928 to 32 in 1963 (Wyz˙ ga, 1993a). According to Schumm (1969), changes in water and/or sediment discharge of a river can be inferred from changes in the morphology of its channel. The increase in sinuosity and depth of the Raba River channel and the reduction in its gradient and meander wavelength clearly testify to the decrease in sediment load of the river over the ﬁrst half of the 20th century (Wyz˙ ga, 2001a). At the same time, the decrease in the width/depth ratio of the channel indicates a reduction in the percentage of bedload in the total river load during that period (Wyz˙ ga, 1993a). The channel bars of the 19th-century Raba River were formed by bimodal, normally loose and overloose gravels (Fig. 20.3a). These channel sediments were massive and very poorly sorted and no armour layers have been observed within them. Such sediments must have been deposited by ﬂood waves with high peak discharges and short time bases that carried huge volumes of bed-material load (Wyz˙ ga, 1993a). The river, heavily overloaded with sediment, transported not only gravel particles but also much sand as bedload. After a peak of such a ﬂood wave, the sediment was deposited rapidly and subjected to little or no reworking. The point bar deposits of the sinuous river from the 1950s were characterized by a highly variable texture. However, overloose gravels now decreased in signiﬁcance in favour of ﬁlled underloose gravels and openwork gravels (Fig. 20.3b) that were lacking in the 19th-century channel. These younger sediments were typiﬁed by a lower content of sand and improved sorting, and their tighter packing was accompanied by the occurrence of armour layers and imbricated gravels. The changes in channel sedimentation that occurred over the ﬁrst half of the 20th century indicate a reduction in the sediment load of the Raba River and a decreasing ﬂashiness of its ﬂood ﬂows during that period.

A review on channel incision in the Polish Carpathian rivers

533

a

Φ

b

12.5

Φ

Figure 20.3. Bar gravels of the middle Raba River and histograms of their grain-size distribution (in weight frequency per cent): (a) massive, underloose gravels from the second half of the 20th century showing bimodal grain-size distribution; (b) unimodal, openwork gravels from the 1950s. Sediment size is expressed in F units. Hatched area in the histogram of sample A represents percentage of undivided ﬁnes.

While the channelization of the middle Raba between 1955 and 1987 signiﬁcantly changed hydraulic conditions in its channel, the tendency to decrease the percentage of sand in the bar deposits continued through the second half of the century and caused a marked reorganization of the facies pattern (Wyz˙ ga, 1993a,b, 2001a). By 1987 overloose gravels ceased to originate in the channel. Instead, deposition of underloose gravels increased in signiﬁcance, accompanied by the formation of bed armouring and numerous pebble clusters on the bed surface. Also in other Carpathian valleys there were observed similar changes in the mode of channel sedimentation leading to a progressive coarsening of bed material, formation of a tightly packed texture and development of bed armouring during the 20th century (cf. Lach and Wyz˙ ga, 2002). Such changes follow a restriction of sediment delivery to channels (e.g., Dietrich et al., 1989; Lisle et al., 1993) and they are typical of rivers recovering from a phase of intense sediment supply (cf. Knighton, 1989; Madej and Ozaki, 1996). A number of alterations in catchment management must have contributed to the reduction in sediment delivery to the Raba and other rivers of the Polish Carpathians during the 20th century (Wyz˙ ga, 1991, 1993a, 2001a). Ploughing along slopes was increasingly replaced by contour ploughing and terracing of cultivated slopes was introduced. The grazing of forest areas, common in the 19th century, ceased after the World War II. The eastern montane section of the Polish Carpathians (upper parts of

534

B. Wyz˙ga

the catchments of the Bia"a, Wis"oka, Wis"ok and San rivers) was dramatically depopulated in the mid-1940s and much arable land was here turned into pasture and meadows or afforested after the World War II (Lach, 1975). For example, in the upper part of the Wis"oka catchment, the forest cover increased from 30% in 1938 to 67% in 1995 (Lach and Wyz˙ ga, 2002) whereas in the upper San catchment, it increased from 48 to 77% between 1937 and 1997 (Kukulak, 2004). The sediment yield of the catchments must have decreased not only due to the change of vegetation cover on hillslopes but also to a general reduction in human activity in the area since the mid-1940s. For example, as cart tracks were abandoned and became overgrown with grass or bushes, they have progressively ceased to function as pathways for rapid evacuation of water and sediment from the hillslopes (Lach and Wyz˙ ga, 2002). Direct human interventions in the Carpathian river beds also were a reason for the reduced delivery of sediment to their channels in the 20th century (Klimek, 1987; Wyz˙ ga, 2001a). The channel regulation paths for the Carpathian tributaries to the Vistula were designed so as to avoid undercutting by the rivers of valley slopes, terraces and fans of tributaries (K˛edzior, 1928); this, together with the stabilization of the channel banks, must have reduced the amount of sediment supplied by lateral erosion. Lining of the banks of mountain streams and construction of check dams in the headwaters, especially intensive in the western part of the Polish Carpathians, must have reduced the delivery of sediment from montane parts of the catchments. Moreover, a few dam reservoirs have been constructed on the Carpathian rivers starting in 1936 (Fig. 20.2). The reservoirs have trapped all the bedload and most of the suspended-sediment load carried from upstream (Łajczak, 1994), thus interrupting the continuity of sediment transport in the rivers and releasing underloaded water to their downstream reaches (cf. Kondolf, 1997). Modiﬁcations of ﬂood ﬂows generated by the montane and foothill parts of Carpathian catchments must have also inﬂuenced the amount of sediment delivered to the rivers and the conditions of its downstream transfer. The existing record of annual maximum discharges of Carpathian rivers since 1921 shows a reduction in peak ﬂows during the last 80 years and such data are presented for the Wis"oka and Skawa Rivers draining, respectively, the eastern and western part of the Polish Carpathians. At the Łabuzie station on the Wis"oka, the mean annual ﬂood (Q2.33) from the years 1956 to 2000 decreased by 31% in comparison with the period 1921–1955 (Table 20.1) whereas at the Wadowice station on the Skawa, the respective reduction in mean annual ﬂood amounted to 8%. With little change in forest cover of the Skawa catchment during the 20th century, the reduction in peak ﬂows of this river most likely reﬂected a change in precipitation pattern induced by modiﬁcations to the atmospheric circulation above central Europe (cf. Kaszewski and Filipiuk, 2003). It has been shown previously (Wyz˙ ga, 2001a) from the analysis of ﬂood discharges of the Raba that a marked reduction in peak ﬂows over the years 1951–1980 was accompanied by an increase in the duration of low ﬂood discharges, the changes reﬂecting a shift from high ﬂash ﬂoods at the beginning of this period to ﬂattened and more prolonged ﬂood waves at its end. The considerably greater reduction in peak ﬂows of the Wis"oka must have reﬂected both the regulatory effect of the reafforestation of the montane part of its catchment, and the change in precipitation pattern (cf. Wyz˙ ga, 1997). With the reduction in ﬂood discharges of

A review on channel incision in the Polish Carpathian rivers

535

Table 20.1. Estimations of the magnitude of ﬂoods (in m3/s) of given recurrence intervals at the Łabuzie and Brzez´nica stations on the Wis"oka River for the record periods 1921–1955 and 1956–2000. Brzez´nica

Łabuzie

Q1.5 Q2.33 Q5 Q10 Q20

1921–1955

1956–2000

1921–1955

1956–2000

385 525 720 880 1035

272 360 575 745 910

370 505 690 850 990

295 410 630 800 960

Note: QX denotes discharge of given recurrence interval. Recurrence intervals of ﬂood discharges were determined from frequency curves graphically ﬁtted to the points plotted on a logarithmic graph paper.

Carpathian rivers, the bed-material yield of their catchments must have been reduced. At the same time, the longer duration of low ﬂood ﬂows facilitated outwashing of ﬁner grains from the channel beds and their downstream transfer. The latter effect must have been especially pronounced below dam reservoirs constructed on the rivers.

5.3.

In-stream gravel mining

Gravel exploitation from the channels of the Carpathian rivers has been a common practice since the 1950s. Especially large amounts of sediment were mined in the ﬁrst two decades after the World War II from the Wis"oka and its tributary, the Ropa (Augustowski, 1968; Osuch, 1968; Rinaldi et al., 2005), which had relatively ﬁnegrained bed material, suitable for concrete production. Between 1941 and 1966, at least 1 million m3 of gravel were mined from a few kilometre-long section of the Ropa in its lower course and the exploitation ended with the complete exhaustion of the gravel resources in the channel in the mid-1960s (Augustowski, 1968). In the years 1955–1964, 2.1 million m3 of sediment were extracted from the channel of the Wis"oka River in its lower and middle course (Osuch, 1968). From that volume, about 1.5 million m3 of sediment were mined from a 30 km-long river reach located where the Wis"oka ﬂows from the Carpathian Foothills onto the foreland basin. The volumes of sediment taken from both rivers were large in comparison to the amounts of bed material stored in the channels and their extraction have exerted a considerable impact on the rivers. The volume of sediment mined from the Ropa was equivalent to a 1 m-thick layer of material that would be removed from the 25 kmlong section of its channel of 40 m width. The extraction of such a volume of sediment from the relatively short channel stretch must have been possible due to an increase in bedload transport rate in the river caused by steepening of the channel bed upstream of the mining site (cf. Kondolf, 1997). The exploitation lowered the channel bed at the mining site by about 1.5 m and in 1966 the river was incised to bedrock here (Augustowski, 1968). Bed degradation was progressing upstream and

536

B. Wyz˙ga

downstream from the mining site, resulting in transformation of a few tens of kilometres-long reach of the Ropa to bedrock boundary conditions by the early 1980s (Rinaldi et al., 2005). Estimation of the sediment transport rate in the Wis"oka indicated that it would take about 500 years to fully replenish the volume of sediment extracted from its channel (Osuch, 1968). In the 30 km-long river reach with the most intensive exploitation, the amount of mined sediment was equivalent to a 0.68 m-thick layer of bed material removed over the pre-mining, bankfull width of the channel (Osuch, 1968). Rapid degradation of the river bed began concurrently with the onset of the sediment mining and about 2.5 m of channel incision occurred in the middle course of the Wis"oka over the second half of the century (Rinaldi et al., 2005). The incision must have reﬂected a variety of causative factors, including an increase in transport capacity of the river caused by the narrowing of its channel by training structures (Wyz˙ ga, 1997, 2001b) as well as the reduction in bed-material delivery from upstream resulting from the land-use changes in the montane part of the catchment (Lach and Wyz˙ ga, 2002) and the gravel mining in the Ropa (Rinaldi et al., 2005). However, two facts emphasize the importance of the sediment mining as a cause of incision of the Wis"oka channel. The greatest extent of incision on the river, both over the whole 20th century and in the second half of that century was concentrated just in the channel section with the most intensive sediment mining (Fig. 20.2 – the Łabuzie and Brzez´nica stations). Moreover, the Wis"oka is characterized by the greatest extent of the 20th-century channel incision among the main rivers of the Polish Carpathians (Fig. 20.2) although all the rivers were similarly channelized in the century and in many of them, excluding the Wis"oka, continuity of sediment transport from the montane parts of their catchments was interrupted by dam reservoirs. Although large-scale gravel mining from the channels of Carpathian rivers has been prohibited since the 1970s, minor amounts of sediment have still been extracted from the rivers in the recent decades. Removing the whole volume of channel bars positioned against undercut concave banks (Fig. 20.4) has repeatedly been undertaken by river engineers in order to keep the rivers on their designed channelization paths (Wyz˙ ga, 2001a). Moreover, illegal extraction of bed material from some mountain watercourses has been a common practice, especially in the western part of the Polish Carpathians (Radecki-Pawlik, 2002). Owing to their repeated occurrence, such activities must have considerably contributed to the sediment deﬁcit in the rivers, although their impact has been dispersed along the length of the channels.

5.4.

Increase in sediment mobility caused by in-channel human disturbances

Contemporary rivers of the Polish Carpathians are characterized by the occurrence of armour and numerous pebble clusters on the bed surface as well as the predominance of close packing of subarmour material. These features reduce sediment mobility (Laronne and Carson, 1976; Reid et al., 1985; Richards and Clifford, 1991) and their undisturbed occurrence in the channels is crucial for protecting bedmaterial particles from entrainment. However, both bed armouring and internal structure of the channel sediments in the rivers have been repeatedly destroyed in the

A review on channel incision in the Polish Carpathian rivers

537

Figure 20.4. Skimming of the gravel bar positioned against an undercut concave bank in the channel of the middle Raba River undertaken in order to eliminate ﬂow conﬁnement and move the thalweg away from the bank.

course of channelization and gravel mining practices undertaken in the last few decades (Wyz˙ ga, 1991, 2001a). The channelization has involved formation of channels with a smooth bed gradient and trapezoid cross-section, with large volumes of bed material being pushed across the channels by bulldozers. With the present-day tendency of Carpathian tributaries to the Vistula to meander, such a channel form is unstable and the practice has been periodically repeated by river engineers to prevent thalweg meandering and to protect concave banks of the regulated channels from erosion. Illegal exploitation of gravel has been typically associated with crossing the channel by vehicles that destroys close packing of channel sediments and exposes ﬁner grains to ﬂow. Exploitation of cobbles from a bar surface, that removes the protective bed armouring and reduces mean size of channel sediments, has been especially common in the rivers originating in the Tatra massif (Dudziak, 1965). All these factors have facilitated erosion and transportation of bed material, thus increasing its loss to the downstream river reaches and strengthening the tendency to channel incision.

6.

Spatial variation in the course and timing of channel incision

Considerable differences in the course and timing of channel incision during the 20th century can be identiﬁed among the Carpathian tributaries to the Vistula as well as between their different sections and three aspects of the variation are considered below. First, the course of bed degradation varied along a river channel (Wyz˙ ga, 1991, 1993b). It can be demonstrated by a comparison between changes in minimum annual stage of the Raba at the Proszo´wki gauging station and those at the Gdo´w

538

B. Wyz˙ga

Figure 20.5. Changes in minimum annual stage of the Raba River at the Proszo´wki and Gdo´w gauging stations, the Wis"oka River at the Łabuzie station and the Skawa River at the Wadowice station since the beginning of the 20th century. The variation of the changes in minimum annual stage among the stations illustrates differences in the course of channel incision between the low-energy (the Wis"oka at Łabuzie) and high-energy (the Skawa at Wadowice) Carpathian rivers as well as between the low-energy (the Raba at Proszo´wki) and high-energy (the Raba at Gdo´w) sections of a given river.

station located, respectively, in the lower and middle course of the river (Fig. 20.5). In the lower river course, the channel bed degraded at a relatively steady rate. In contrast, in the middle course, channel incision was a result of the series of separate degradation events alternating with periods of vertical stability or aggradation of the channel bed. A similar difference in the course of bed degradation was also recognized between rivers draining the eastern and western part of the Polish Carpathians (Wyz˙ ga, 2001b) and it is shown here for the Łabuzie station on the Wis"oka and the Wadowice station on the Skawa (Fig. 20.5), both gauging stations being located at about one third of the total length of each river upstream of its mouth to the Vistula (Fig. 20.2). At Łabuzie downcutting of the Wis"oka channel proceeded in the second half of the 20th century at a practically constant rate whereas at Wadowice, three long periods of incision of the Skawa channel, separated by short episodes of bed aggradation, are evident over the century.

A review on channel incision in the Polish Carpathian rivers

539

The mentioned variation in the course of channel incision most likely reﬂects differences in channel gradient and, hence, the energy of ﬂood ﬂows between particular rivers or between particular reaches of a given river. In the lower, foreland reaches, Carpathian rivers have more gentle channel gradient and are typiﬁed by lower values of unit stream power at ﬂood ﬂows of a given frequency than in their middle, foothill reaches. Considerable differences in channel gradient and ﬂow energy exist also between the rivers draining the eastern and western part of the Polish Carpathians (Klimek, 1979; Wyz˙ ga, 2001b), conditioned by the differences in the orography of the catchments in both regions, and in the distance between the river headwaters and their recipient, the Vistula, ﬂowing obliquely to the mountains (Fig. 20.2). In their lower reaches as well as in the eastern region, Carpathian rivers have had insufﬁcient energy to destroy the river-control structures and have generally remained laterally stable following their channelization. Under such conditions, the excess energy of the straightened and narrowed rivers has been dissipated by the scouring of their channel beds, this leading to the formation of deep channels bounded by the highly elevated ﬂoodplains as at the Łabuzie station on the Wis"oka (Fig. 20.6a) and the Proszo´wki station on the Raba. In the higher river reaches and in the rivers from the western part of the Polish Carpathians, the relatively high energy of ﬂood ﬂows has facilitated exceedance of the threshold of stability of the channelization structures. With the increase in channel width and sinuosity following a destruction of bank-protection structures, transport capacity of the rivers had been reduced and degradation of their beds had been arrested until the next channelization that again straightened and narrowed the channels. Such alternating periods of incision of the channelized rivers and of lateral channel migration have led to the development of incised meander belts with progressively lower ﬂoodplain levels formed along the incised channels, and to the transformation of the former ﬂoodplains into terraces as at the Wadowice station on the Skawa (Fig. 20.6b) and the Gdo´w station on the Raba. Second, the main phase of incision of Carpathian rivers occurred progressively later in the upstream direction. It is demonstrated by changes in minimum annual stage of the Wis"oka at the Mielec and Z˙ o´"ko´w gauging stations located, respectively, in the lower and upper course of the river (Fig. 20.2). At Mielec about two thirds of the total bed degradation were accomplished over the ﬁrst half of the 20th century ˙ o´"ko´w, the major phase of channel incision took place in the whereas at Z 1960s–1980s (Fig. 20.7). Several factors must have contributed to the variation in timing of the main phase of channel incision observed along the course of Carpathian rivers. Time-shifting of the most intense channelization works for different river reaches caused that the resultant increase in transport capacity of the watercourses occurred ﬁrst in their foreland reaches and later in the foothill and montane reaches (Wyz˙ ga, 2001a; Zawiejska and Wyz˙ ga, 2005). Moreover, steepening of the channel gradient immediately above already incised reaches has induced upstream-progressing bed degradation (Galay, 1983; Wyz˙ ga, 1993b). With the channel width and sinuosity of Carpathian rivers ﬁxed by channelization structures, the increase in channel depth and gradient ﬂattening, occasioned by backward erosion, have been important mechanisms of re-attaining equilibrium conditions in the rivers disturbed by channelization (Wyz˙ ga, 1993b). Finally, in the second half of the 20th century,

B. Wyz˙ga

540

a

b

Figure 20.6. Downstream view of the cross-sections of (a) the Wis"oka River at the Łabuzie gauging station in 1970 and 1996, and (b) the Skawa River at the Wadowice station in 1959 and 1997. Grey-shaded columns indicate a lowering of the stage attained at the ﬂood discharges of given recurrence intervals, that accompanied incision of the channels between the years considered. Elevation of the bankfull stage, hbf, is also marked. Morphological zones of the cross-section: (CH) channel; (FP) ﬂoodplain; (TR) terrace.

montane reaches of the rivers, especially those draining the eastern part of the Polish Carpathians, experienced considerable reduction in sediment delivery in response to the land-use changes in the catchments (Lach and Wyz˙ ga, 2002; Kukulak, 2003). Third, with rapid incision of a major tributary, the cessation or slowing down of bed degradation in the stem river was observed (Wyz˙ ga, 1993c). Such a situation occurred in the lower section of the San where rapid channel incision, recorded in the ﬁrst half of the 20th century, has ceased since the 1950s with the onset of intensive

A review on channel incision in the Polish Carpathian rivers

541

Figure 20.7. Changes in minimum annual stage of the Wis"oka at the Mielec gauging station in the lower ˙ o´"ko´w station in its upper course since 1890. course of the river and at the Z

downcutting of the Wis"ok channel (Fig. 20.2 – see the Nisko and Rzucho´w stations). At the same time, in the reach of the San located upstream of the Wis"ok mouth, rapid channel incision continued in the second half of the century (Fig. 20.2 – see the Jaros"aw station). Apparently, a plentiful supply of bed material having been ﬂushed out from the deepening Wis"ok channel must have largely promoted the vertical stabilization of the lower San bed. The different tendencies of the vertical position of the San channel observed in the second half of the century downstream and upstream of the tributary indicate that with transport capacity of this and other Carpathian rivers considerably increased owing to their channelization, stabilization of vertical position of the river beds was only possible with an abundant delivery of bed material.

7.

Importance of the main incision drivers

A comparison of the amount and timing of incision of Carpathian rivers with the location and time of occurrence of its major drivers enables determining the relative importance of particular factors for producing bed degradation in the watercourses. At the regional scale, river channelization was undoubtedly the most important reason for the channel incision. It disturbed vertical stability of the rivers by increasing considerably their transport capacity, reducing sediment delivery to the channels through their disconnection from valley slopes and bed material stored on the valley ﬂoors and increasing susceptibility of bed-material particles to entrainment due to a frequent disturbance of the channel beds in the course of repeated rivercontrol works. The commencement of bed degradation in the Carpathian rivers almost concurrently with the beginning of major regulation works in their channels in 1904 (Wyz˙ ga, 1991, 2001a), the progressive shifting in the upstream direction of

542

B. Wyz˙ga

both the most intensive changes to river planform and the main phase of channel incision (Wyz˙ ga, 2001a; Zawiejska and Wyz˙ ga, 2005) as well as a clear relation between the degree of channel narrowing and the amount of bed degradation along a river course (Wyz˙ ga, 1993b) point to channelization works as the principal cause of the degradational tendency of Carpathian rivers during the 20th century. The increase in transport capacity of the rivers caused by their channelization was so large that even with a delivery of bed material ﬂushed out from the deepening upstream reaches, the channels have not been able to recover to their pre-disturbance vertical position. Large-scale gravel mining carried out in some Carpathian rivers during the 1940s–1960s greatly increased sediment deﬁcit as well as the resultant rates and total extent of bed degradation in their channels. In the river reaches subject to such activity, either the greatest channel incision among the Carpathian rivers (3.8 m in the lower course of the Wis"oka and up to 3.5 m in the Czarny Dunajec) or transformation from alluvial to bedrock boundary conditions (the Ropa River) were recorded (Rinaldi et al., 2005; Zawiejska and Wyz˙ ga, 2005). Among direct interventions in the Carpathian rivers, construction of dam reservoirs appears to have exerted the least inﬂuence on the hitherto accomplished incision of their channels. This situation reﬂects a small number of the reservoirs, their construction relatively late in the 20th century (e.g., compare the impoundment of the Raba River at Dobczyce in 1987 with the course of channel incision at the downstream located Gdo´w gauging station – Fig. 20.5) and/or location of some of them in the upper river reaches (the reservoirs on the San, Wis"ok and Ropa Rivers) where they have trapped sediment delivered from only a small proportion of the total area of the catchments. Only the Pora˛ bka Dam on the So"a and the Roz˙ no´w Dam on the Dunajec, constructed in the late 1930s at the half-length of the river courses (Fig. 20.2), must have signiﬁcantly affected degradational processes in the downstream reaches of these rivers. The land use changes that occurred in the eastern montane part of the Polish Carpathians during the second half of the 20th century must have substantially and persistently reduced sediment delivery to the rivers draining that area. This is shown by the widespread transformation of the former alluvial channels into bedrock ones over recent decades (Lach and Wyz˙ ga, 2002; Kukulak, 2003), also in the river reaches within forested corridors where their channels have not been regulated. In valley sections underlain by a thick cover of alluvium, a combined operation of the increase in transport capacity of the rivers caused by their channelization and the reduction in sediment supply following the catchment reafforestation has led to deep, rapid channel incision (see the Z˙ o´"ko´w station – Figs. 20.7 and 20.8a) (Lach and Wyz˙ ga, 2002). However, it should be noted that with little proportion of mountain areas in the catchments of the rivers draining the eastern part of the Polish Carpathians (Fig. 20.2), the impact of the hillslope reafforestation must have progressively declined in the downstream direction. In the western part of the Polish Carpathians, land-use changes took place gradually since the last decades of the 19th century and were mostly represented by alterations in some agricultural practices (Wyz˙ ga, 1991, 2001a), with little modiﬁcation to forest cover of the area. Although the resultant reduction in sediment yield of the catchments must have been considerably lower

A review on channel incision in the Polish Carpathian rivers

543

than in the eastern part of the mountains, its impact on the river channels is shown by the transformation of the braided channel of the 19th-century middle Raba into the sinuous channel in the mid-20th century. With the decreased sediment delivery from the catchments, channelization works carried out in the second half of the 20th century caused a greater disturbance to vertical stability of the rivers than the works from the early decades of the century (Wyz˙ ga, 2001a).

8.

Effects of the channel incision

The deep incision of the rivers of the Polish Carpathians during the 20th century resulted in unintentional effects in their channels and on the valley ﬂoors, which are unfavourable for the natural environment and the economy (Froehlich, 1980; Klimek, 1983; Wyz˙ ga, 1991, 2001a). The undermining of bridge piers and regulation structures is the reason for expensive repairs. With water stage in the rivers lowered below the level of water intakes, the construction of weirs is necessary to enable their further operation. The lowering of water table on the valley ﬂoors, which follows a fall of medium and low stages in the rivers, causes several detrimental effects. These include: (i) a loss of groundwater resources; (i) drying of the soil of cultivated land on the valley ﬂoors, reducing root-crop yields; and (iii) drying up of oxbow lakes and the impoverishment of plant and animal communities of riverside ecosystems. Finally, the lowering of water level in the rivers below a dense network of roots of riparian vegetation facilitates undermining and fast retreat of the channel banks. Such adverse effects are apparent at a local scale, but of primary signiﬁcance seems to be that the potential of the ﬂoodplains of the Carpathian rivers for water and sediment storage has been dramatically reduced with the progress of channel incision. A considerable increase in the ﬂood hazard to downstream reaches of the Carpathian rivers has been recorded following their channelization and subsequent incision (Wyz˙ ga, 1996, 1997). It is shown by a comparison of the magnitude of ﬂoods of given recurrence intervals at the Łabuzie and Brzez´nica stations on the Wis"oka, estimated for the record periods 1921–1955 and 1956–2000 characterized, respectively, by small and high degree of incision of the river channel (Table 20.1). With its length of 21 km, the Łabuzie-Brzez´nica reach is long enough to allow signiﬁcant ﬂood wave transformation. As the catchment area increases along the reach by only 12%, recorded trends of ﬂood magnitude cannot be attributed to changed inﬂow from tributaries. At both stations, in the second period ﬂood magnitudes decreased in comparison with those recorded in the ﬁrst period (Table 20.1) in response to reafforestation of the montane part of the catchment (Lach, Wyz˙ ga, 2002) and, probably, also to some change in precipitation pattern (see Part 5.2). However, a scale of the reduction at both stations was distinct, with mean annual ﬂood lowered by 31% at Łabuzie and 19% at Brzez´nica, and the difference can be attributed to changed conditions of ﬂood wave transformation between the stations. While in the years 1921–1955 the magnitude of all considered index ﬂoods (Q1.5–Q20) slightly decreased in the reach, after 1955 the increase in peak discharges of the ﬂood waves passing the reach was apparent, reaching from 5% for a 20-year ﬂood to 14% for mean annual ﬂood (Table 20.1).

544

B. Wyz˙ga

Also for the Gdo´w-Proszo´wki reach of the Raba River, a considerable increase over time of the peak discharges recorded at its downstream end has been recognized by comparing the annual maximum discharges from both stations as well as the inﬂow and outﬂow peak discharges for all ﬂood waves from particular decades (Wyz˙ ga, 1996, 1997). A remarkable coincidence of the channel incision in the Łabuzie-Brzez´nica reach of the Wis"oka, the Gdo´w-Proszo´wki reach of the Raba and in some other reaches of Carpathian rivers, and of the increase in peak discharges recorded at their downstream ends indicates a link between the two phenomena (Wyz˙ ga, 1996, 1997). As ﬂood ﬂows were increasingly concentrated in the deepened channels with the advancing incision, water retention in the ﬂoodplain areas was progressively reduced. Moreover, ﬂows conveyed in the progressively deeper channels were characterized by increasingly high relative smoothness of ﬂow (ratio of water depth to the height of protrusion of bed-material particles to the ﬂow). This, together with the reduction in channel-form resistance resulting from the channelization, must have also decreased attenuation of in-channel ﬂood waves. A reduction of ﬂood hazard was one of the main aims of the channelization of Carpathian rivers; however, it had an opposite effect. Although ﬂood stages lowered in the upper parts of the incised river reaches, the danger has been merely shifted downstream and, at the same time, magniﬁed there owing to the increasingly peaked nature of ﬂood waves passing the incised reaches. The increase in the ﬂood hazard to downstream river reaches has been commonly overlooked by river managers owing to the simultaneous reduction in ﬂood ﬂows generated by the montane parts of the Carpathian catchments. However, future consequences of the increase are likely to be diverse in the rivers draining the eastern and western part of the Polish Carpathians (Wyz˙ ga, 1997). In the rivers from the eastern part, the lowering of ﬂoods formed in the past few decades in the montane parts of their catchments has mainly resulted from the reafforestation of these areas and, thus, is likely to continue in the future. The magniﬁcation of peak ﬂows in the incised reaches of these rivers will simply mean the lost chance for permanent reduction of ﬂood hazard on the watercourses that could be achieved as a result of the land-use change. On the contrary, the occurrence of meteorologic conditions that in the past few decades favoured the formation of lower ﬂood ﬂows on the rivers in the western part of the mountains seems temporary. Here, the recovery of heavy rains in subsequent years would suddenly manifest the potential ﬂood hazard resulting from the channel incision. Incision of the Carpathian rivers has also considerably decreased the potential of their ﬂoodplains for sediment storage (Wyz˙ ga, 2001b). However, different factors have played a dominant role in reducing the depositional potential of the ﬂoodplain ﬂows in the valley reaches with an incised channel, typical of the Carpathian rivers with lower energy, and in those with an incised meander belt, characteristic of the high-energy rivers. This was shown by analysing the changes in ﬂow characteristics of the Wis"oka River at the Łabuzie station between 1970 and 1996, as well as the Skawa River at the Wadowice station between 1959 and 1997, with minimum annual stage of both rivers lowered during the periods by a similar value of 1–1.2 m (Wyz˙ ga, 2001b). At Łabuzie, substantial lowering of stages for low ﬂood discharges contrasted with markedly smaller one for high-magnitude ﬂoods. For instance, the stage associated with mean annual ﬂood lowered in the years 1970–1996 by 110 cm

A review on channel incision in the Polish Carpathian rivers

545

whereas that attained at a 20-year ﬂood fell by 28 cm (Fig. 20.6a). At Wadowice, the concentration of ﬂood water in the deepened channel and over the new, low-lying ﬂoodplain caused a substantial lowering of stage both for low and high ﬂood discharges. In the years 1959–1997 it amounted to 160 cm for mean annual ﬂood and 94 cm for the discharge with a 20-year return period (Fig. 20.6b). With this large fall of ﬂood stages of the river, its former ﬂoodplain has become transformed into a terrace, and the lateral extent of ﬂood water on the valley ﬂoor has shrunk considerably (Fig. 20.6b). Linked with the reduced vertical and lateral extent of ﬂood water in the river cross-sections was also a reduction in the frequency and duration of submergence of particular levels on the valley ﬂoors at given ﬂood discharges. On the Skawa River, the reduction was considerably greater than on the Wis"oka (Wyz˙ ga, 2001b). The amount of ﬂood water conveyed in the extra-channel zone of the rivers was reduced with the increased concentration of ﬂood ﬂows in the river cross-sections (Table 20.2). For major ﬂoods, the scale of the reduction was apparently greater on the Skawa than on the Wis"oka River (Table 20.2). This may be explained by the retained channel width of the Skawa with the progress of river incision and the constriction of both channel and ﬂoodplain ﬂows in the river cross-section (Fig. 20.6b). Where bed degradation was associated with lateral channel stability, it has increased the relative elevation of ﬂood stages above the channel bed (Wyz˙ ga, 2001b). At Łabuzie on the Wis"oka River, the elevation above the mean bed level of the stage attained at the 5-year ﬂood increased from 6.7 m in 1970 to 8.31 m in 1996 and that of the stage of the 20-year ﬂood increased from 8.11 to 10.13 m. With the increased distance between river bed and water surface and the decreased depth of ﬂoodplain inundation at particular discharges (Fig. 20.6a), at present, the river banks can be overtopped only by the uppermost parts of ﬂood water. Moreover, vertical differentiation of the size and concentration of sediment particles transported in graded suspension must have increased with the greater depth of the water column. The operation of these factors tended to reduce the amount of coarse sediment particles

Table 20.2. Percentage of the total ﬂow of the Wis"oka River at the Łabuzie station, and of the Skawa River at the Wadowice station, conveyed in the extra-channel zone of the gauging sections before and after the period of rapid incision of both rivers.

Q3 Q5 Q10 Q15 Q20

Wis"oka River at Łabuzie

Skawa River at Wadowice

1970

1996

1959

1997

2.7 7.3 12.3 15.3 17.4

0.1 1.2 5.2 8.2 10.3

1.3 5.5 12.5 16.3 19.0

0.2 1.6 4.2 5.7 6.7

Note: QX denotes discharge of given recurrence interval. On the Wis"oka, the extra-channel zone is equivalent to the ﬂoodplain but on the Skawa, it comprises the ﬂoodplain and the lower terraces.

546

B. Wyz˙ga

introduced with ﬂood water onto the ﬂoodplain, thus increasing concentration of suspended-sediment transport within the incised channel. Changes in ﬂow velocity have been another effect of incision of the Carpathian rivers that must have affected overbank deposition on the valley ﬂoors. Mean velocity in channel zone typically increased with the progress of bed degradation but the direction of change of the velocity in extra-channel zone differed between the valley reaches with an incised channel and those with an incised meander belt (Wyz˙ ga, 2001b). At Łabuzie on the Wis"oka River, mean velocity in the ﬂoodplain zone decreased between 1970 and 1996 with the reduced depth of ﬂoodplain submergence at particular discharges. At Wadowice on the Skawa, in 1997 mean velocity over the low-lying ﬂoodplain was 20–30% higher than velocity typifying the ﬂoodplain ﬂows in 1959. Still greater increase in ﬂow velocity accompanied the formation of an incised meander belt of the upper Wis"oka River at Z˙ o´"ko´w, which originated through dissection of the former, wider channel bed by about 2.5 m (Fig. 20.8a) (Lach and Wyz˙ ga, 2002). Here, mean velocity in the channel zone increased between 1963 and 1991 by 35–45% but mean velocity in the extra-channel zone is now three to four times faster than before channel incision (Fig. 20.8b). Investigations carried out in the Raba valley have identiﬁed a dramatic reduction in the rate of overbank sediment accretion that followed channelization and the subsequent incision of this river (Wyz˙ ga, 1991). Field observations on the Skawa and Wis"oka Rivers conﬁrm an insigniﬁcant role played nowadays by ﬂoodplain sedimentation in the valleys of Carpathian rivers. At the peak discharge of the ﬂood of July 1997 on the Skawa River at Wadowice, which had a 29-year recurrence interval, mean velocity of ﬂow over the narrow, low-lying ﬂoodplain attained 1.68 m/s. Although bankfull stage was exceeded for more than 2 days in this section and the ﬂoodplain was covered by about 2 m of water at the ﬂood peak, the thickness of ﬂoodplain deposition was relatively low, varying from 1.5 to 7 cm. At the same time, conspicuous levee sediments ranging up to 30 cm in thickness were deposited during the ﬂood on the Vistula River ﬂoodplain immediately downstream of the junction with the Skawa. Here, the Vistula is not incised and mean velocity of the ﬂoodplain ﬂow at the ﬂood peak was estimated at 0.3 m/s (Wyz˙ ga, 1999). These contrasting hydraulic and depositional patterns show that at present, during major ﬂoods carrying large sediment loads, ﬂows over the narrow ﬂoodplains formed along the incised channels of Carpathian rivers are too fast to allow signiﬁcant overbank deposition (Wyz˙ ga, 2001b). In turn, with the contrast in mean velocity between the channel and ﬂoodplain ﬂows of the Wis"oka River at Łabuzie increased owing to channel incision, large amounts of the coarser sediment carried in suspension into the ﬂoodplain area should be deposited on the natural levees. However, despite the occurrence of two major ﬂoods in 1987 and 1989, which had recurrence intervals of 35 and 18 years, respectively, no prominent levees have been observed along the incised channel of this river. Apparently, little sediment is carried by the upper parts of the water column overtopping the river banks, even at such high ﬂoods, whereas the transport of coarser fractions of the suspended load takes place almost entirely within the incised channel (Wyz˙ ga, 2001b). Table 20.3 summarizes major factors which have affected overbank deposition on the rivers of the Polish Carpathians in response to their incision, operating in the

A review on channel incision in the Polish Carpathian rivers

547

Figure 20.8. (a) Downstream view of the cross-section of the Wis"oka River at the Z˙ o´"ko´w gauging station in 1963 and 1991. Morphological zones of the cross-section: (CH) channel; (FP) ﬂoodplain; (TR) terrace. Elevation of the bankfull stage, hbf, in 1963 and 1991 is also marked. (b) Relationship between the mean ﬂow velocity in total cross-section (1, 1a), in channel zone (2, 2a) and in extra-channel zone (3, 3a) of the cross-section, and discharge for the Z˙ o´"ko´w station on the Wis"oka River in 1963 (1–3) and 1991 (1a–3a). Discharges are referred to their recurrence interval determined by the annual maximum series method from the years 1951 to 2000.

B. Wyz˙ga

548

valley reaches with an incised channel and in those with an incised meander belt. Despite the different combination of the factors in both valley types, their overall result has been a considerable reduction of the potential of the Carpathian ﬂoodplains for sediment storage during the past few decades (Wyz˙ ga, 2001b). Consequently, the majority of the suspended load of the rivers may now be routed through their incised reaches directly to the Vistula, contributing to the rapid channel and ﬂoodplain aggradation in the middle course of that river (Łajczak, 1997).

9.

Concluding remarks

Up to 3.8 m of channel incision occurred over the 20th century in the rivers of the Polish Carpathians. With its extent and rate, the incision has undoubtedly been a spectacular phenomenon in the Holocene history of the rivers. It resulted from a combination of factors which increased transport capacity of the rivers while reducing availability of bed-material calibre sediments for ﬂuvial transport. The earlier occurrence of a main phase of channel incision in the lower sections of Carpathian rivers and later in their higher sections indicates the channelization-induced increase in transport capacity of the watercourses as the principal cause of bed degradation at the regional scale, with the reduction in catchment sediment supply being a less important or later operating factor. In some of the rivers, in-stream gravel mining has additionally increased the deﬁcit of sediment available for ﬂuvial transport. A similar history of channel changes has also been recorded for mountain and piedmont rivers of western (Bravard et al., 1997; Landon et al., 1998) and southern Europe (Rinaldi, 2003; Surian and Rinaldi, 2003). The tendency towards channel aggradation that typiﬁed the rivers from these regions during the Little Ice Age, has been reversed since around the beginning of the 20th century as a result of river-control works and land use changes in the catchments. In-stream gravel mining and construction of dam reservoirs have dramatically increased bed degradation on the rivers since about the mid-20th century. Hydroclimatic changes after the end of the Little Ice Age have been reported as one of the reasons for the reduced sediment supply to the rivers of western Europe (Bravard et al., 1997) and the existing evidence for the modiﬁed pattern of atmospheric circulation over central Europe allows also to invoke changes in precipitation pattern as a cause of the reduced sediment yield of the Carpathian catchments during the 20th century. However, the distant location of the regions, with their mountain ranges differently oriented in relation to the general directions of atmospheric circulation over the northern hemisphere, casts doubt on the climatic changes as a major causative factor in reducing sediment yield of the mountain catchments in Europe. Rather, the equivalent evolution of the rivers from different regions reﬂects a similar history of the changes in channel and catchment management. With a variety of the identiﬁed detrimental effects of channel incision, changes in management of the Carpathian rivers and the degrading rivers of the mentioned regions are necessary to arrest degradation of their beds and re-establish the conditions for water and sediment storage on the ﬂoodplains. Any human activities in channels, that cause the removal of bed material from the rivers or facilitate its loss to downstream reaches, should be stopped (Wyz˙ ga, 2001a; Bojarski et al., 2005;

Valley reaches with an incised river channel

Valley reaches with an incised meander belt

Frequency and duration of the valley ﬂoor submergence at a given ﬂood discharge

Decreased

Percentage of the total ﬂow conveyed in the extra-channel zone

Decreased both for low and high-ﬂood discharges

Relative elevation of ﬂood stages above the channel bed (affecting concentration of suspended sediment transport within the channel) Flow velocity in the extra-channel zone

Increased

Slightly decreased for the new, low-lying ﬂoodplain but considerably decreased for the former ﬂoodplain Decreased for low ﬂood discharges, considerably decreased for high-ﬂood discharges Unchanged or slightly decreased

Decreased

Increased over the new, low-lying ﬂoodplain but considerably decreased over the former ﬂoodplain

A review on channel incision in the Polish Carpathian rivers

Table 20.3. Factors affecting overbank deposition on the rivers of the Polish Carpathians in response to channel incision, shown for the valley reaches with an incised river channel and those with an incised meander belt.

549

550

B. Wyz˙ga

Rinaldi et al., 2005). Thus, extracting gravel from channels, practiced by river engineers to keep the rivers on their channelization paths, must be abandoned and uncontrolled sediment mining should be rigorously forbidden. River-control works must be carried on in such a way that minimizes or eliminates the destruction of bed armouring and the internal structure of channel sediments. Most importantly, a reduction in transport capacity of the rivers is necessary to compensate for their present-day decreased sediment loads (Wyz˙ ga, 2001a). In mountain valleys, it could be achieved with some widening of the river channels but in foothill and foreland reaches, where the tendency to river meandering reappeared in the 20th century, an increase in channel sinuosity should be allowed wherever it is possible without an erosional threat to property and infrastructure. The active meandering would reduce channel gradient and, hence, transport capacity of the rivers, whereas the input of material from lateral erosion would help to increase sediment loads not only in the reaches where active meandering is allowed but also in those in which channelization schemes are maintained. Recently, a proposal for a new management policy of the streams and rivers from the Carpathian part of the upper Vistula drainage basin has been formulated (Bojarski et al., 2005), in which various methods of restoring the disturbed equilibrium conditions in mountain watercourses are presented and discussed in terms of their usefulness under different styles of valley ﬂoor management. For narrow streams in forested corridors, it is suggested to allow spontaneous formation of wood dams from fallen trees. As wood dams facilitate dissipation of the energy of ﬂood ﬂows and force in-channel sediment storage, mountain streams with abundant wood accumulations can maintain their alluvial beds even under relatively steep valley gradients (Montgomery et al., 1996). For wider watercourses ﬂowing far from settlements and infrastructure, free channel migration within erodible river corridors should be allowed (cf. Pie´gay et al., 2005), with anti-erosion revetments located at the boundaries of the ﬂoodplain area. Where planform stability of the channelized stream or river in an urbanized area must be preserved, construction of artiﬁcially elevated rifﬂes made of layers of compacted boulders is recommended to reduce the excessive ﬂow capacity of the incised channel.

Acknowledgements Free access to unpublished data of the Hydrologic Survey is kindly acknowledged. Thanks are also due to two anonymous referees for their helpful criticism of the manuscript.

References Augustowski, B., 1968. Spostrzez˙ enia nad zmianami antropogenicznymi w korycie rzeki Ropy w Karpatach (Observations on man-caused changes in the channel of the River Ropa in the Carpathians). Zesz. Geogr. WSP w Gdan´sku 10, 161–168 (in Polish, with English summary).

A review on channel incision in the Polish Carpathian rivers

551

Bathurst, J.C., 1982. Theoretical aspects of ﬂow resistance. In: Hey, R.D., Bathurst, J.C., and Thorne, C.R. (Eds), Gravel-Bed Rivers. Wiley, Chichester, pp. 83–108. Bojarski, A., Jelen´ski, J., Jelonek, M., et al., 2005. Zasady dobrej praktyki w utrzymaniu rzek i potoko´w go´rskich (Good-practice manual of sustainable maintenance of mountain streams and rivers in southern Poland). Ministerstwo S´rodowiska, Warszawa (in Polish, with English summary). Bravard, J.P., Amoros, C., Pautou, G., et al., 1997. River incision in South-east France: morphological phenomena and ecological effects. Regul. Rivers Res. Manage. 13, 1–16. Brookes, A., 1987. River channel adjustment downstream from channelization works in England and Wales. Earth Surf. Process. Landf. 12, 337–351. Bull, W.B., Scott, K.M., 1974. Impact of mining gravel from urban stream beds in the southwestern United States. Geology 2, 171–174. Church, M., 1978. Palaeohydrological reconstructions from a Holocene valley ﬁll. In: Miall, A.D. (Ed.), Fluvial Sedimentology. Can. Soc. Petrol. Geol. Memoir, Vol. 5, pp. 743–772. Collins, B., Dunne, T., 1989. Gravel transport, gravel harvesting, and channel-bed degradation in rivers draining the southern Olympic Mountains, Washington, USA. Environ. Geol. Water Sci. 13 (3), 213–224. Dietrich, W.E., Kirchner, J.W., Ikeda, H., Iseya, F., 1989. Sediment supply and the development of the coarse surface layer in gravel-bedded rivers. Nature 340, 215–217. Dudziak, J., 1965. Dzika eksploatacja kamienia w powiecie nowotarskim (Uncontrolled exploitation of rocks in the district of Nowy Targ). Ochrona Przyrody 31, 161–187. Dynowska, I., 1991. Bilans wodny. In: Dynowska, I. and Maciejewski, M. (Eds), Dorzecze go´rnej Wis"y. PWN, Warszawa-Krako´w, pp. 223–227. Emerson, J.W., 1971. Channelization: a case study. Science 173, 325–326. Froehlich, W., 1980. Hydrologiczne aspekty pog"˛ebiania koryt rzek beskidzkich (Deepening of stream channels in the Beskidy Mts – a hydrological aspect). Zeszyty Problemowe Post˛epo´w Nauk Rolniczych 235, 257–268 (in Polish, with English summary). Froehlich, W., 1982. Mechanizm transportu ﬂuwialnego i dostawy zwietrzelin do koryta w go´rskiej zlewni ﬂiszowej (The mechanism of ﬂuvial transport and waste supply into the stream channel in a mountainous ﬂysch catchment). Pr. Geogr. Instytutu Geograﬁi PAN 143, 1–144 (in Polish, with English summary). Galay, V.J., 1983. Causes of river bed degradation. Water Resour. Res. 19, 1057–1090. Kaszewski, B.M., Filipiuk, E., 2003. Variability of atmospheric circulation in Central Europe in the summer season 1881–1998 (on the basis of the Hess-Brezowski classiﬁcation). Meteorologische Zeitschrift 12 (3), 123–130. K˛edzior, A., 1928. Roboty wodne i melioracyjne w Po"udniowej Ma"opolsce wykonane z inicjatywy Sejmu i Wydzia"u Krajowego, Lwo´w. Klimek, K., 1979. Geomorfologiczne zro´z˙ nicowania koryt karpackich dop"ywo´w Wis"y (Morphodynamic channel types of the Carpathian tributaries to the Vistula). Folia Geogr. Ser. Geogr. Phys. 12, 35–47 (in Polish, with English summary). Klimek, K., 1983. Erozja wg"˛ebna dop"ywo´w Wis"y na przedpolu Karpat (Vertical erosion of Vistula tributaries on the Carpathian foreland). In: Kajak, Z. (Ed.), Ekologiczne podstawy zagospodarowania Wis"y i jej dorzecza, PWN, Warszawa-Ło´dz´, pp. 97–108 (in Polish, with English summary). Klimek, K., 1987. Man’s impact on ﬂuvial processes in the Polish Western Carpathians. Geogr. Ann. 74A, 123–131. Klimek, K., Trafas, K., 1972. Young-Holocene changes in the course of the Dunajec River in the Beskid Sa˛ decki Mts (Western Carpathians). Stud. Geomorph. Carp.-Balcan. 6, 85–92. Knighton, A.D., 1989. River adjustment to changes in sediment load: the effects of tin mining on the Ringarooma River, Tasmania, 1875–1984. Earth Surf. Process. Landf. 14, 333–359. Kondolf, G.M., 1997. Hungry water: effects of dams and gravel mining on river channels. Environ. Manage. 21, 533–551. Krzemien´, K., 1981. Zmiennos´ c´ subsystemu korytowego Czarnego Dunajca (The changeability of the Czarny Dunajec channel subsystem). Zesz. Nauk. Uniwersytetu Jagiellon´skiego, Pr. Geogr. Vol. 53, pp. 123–137 (in Polish, with English summary).

552

B. Wyz˙ga

Krzemien´, K., 2003. The Czarny Dunajec River, Poland, as an example of human-induced development tendencies in a mountain river channel. Landf. Anal. 4, 57–64. Kukulak, J., 2003. Ewolucja koryta Sanu w Bieszczadach Wysokich w drugiej po"owie minionego 1000-lecia (Evolution of the San River channel in the Bieszczady Wysokie Mts in the second half of the last millennium). In: Lach, J. (Ed.), Dynamika zmian s´ rodowiska geograﬁcznego pod wp"ywem antropopresji. Akademia Pedagogiczna, Krako´w, pp. 134–140 (in Polish, with English summary). Kukulak, J., 2004. Zapis skutko´w osadnictwa i gospodarki rolnej w osadach rzeki go´rskiej na przyk"adzie aluwio´w dorzecza go´rnego Sanu w Bieszczadach Wysokich (Record of human settlement and agriculture in mountain river alluvium: the case of the upper San River in the Bieszczady Wysokie). Pr. Monogr. Akademii Pedagogicznej w Krakowie, Vol. 381, pp. 1–125 (in Polish, with English summary). Lach, J., 1975. Ewolucja i typologia krajobrazu Beskidu Niskiego z uwzgl˛ednieniem gospodarczej dzia"alnos´ ci cz"owieka (Evolution and typology of the landscape in the Low Beskid Mts with special consideration of the economic activity of man). Pr. Monogr. WSP w Krakowie, Vol. 16, pp. 5–72 (in Polish, with English summary). Lach, J., Wyz˙ ga, B., 2002. Channel incision and ﬂow increase of the upper Wis"oka River, southern Poland, subsequent to the reafforestation of its catchment. Earth Surf. Process. Landf. 27, 445–462. Łajczak, A., 1994. The rates of silting and the useful lifetime of dam reservoirs in the Polish Carpathians. Z. Geomorf. 38, 129–150. Łajczak, A., 1997. Anthropogenic changes in the suspended load transportation by and sedimentation rates of the River Vistula, Poland. Geogr. Polon. 68, 7–30. Landon, N., Pie´gay, H., Bravard, J.P., 1998. The Droˆme river incision (France): from assessment to management. Landsc. Urban Plan. 43, 119–131. Laronne, J.B., Carson, M.A., 1976. Interrelationships between bed morphology and bed-material transport for a small, gravel-bed channel. Sedimentology 23, 67–85. Lie´bault, F., Pie´gay, H., 2001. Assessment of channel changes due to long-term bedload supply decrease, Roubion River, France. Geomorphology 36, 167–186. Lisle, T.E., Iseya, F., Ikeda, H., 1993. Response of a channel with alternate bars to a decrease in supply of mixed-size bed load: a ﬂume experiment. Water Resour. Res. 29, 3623–3629. Madej, M.A., Ozaki, V., 1996. Channel response to sediment wave propagation and movement, Redwood Creek, California, USA. Earth Surf. Process. Landf. 21, 911–927. Montgomery, D.R., Abbe, T.B., Bufﬁngton, J.M., et al., 1996. Distribution of bedrock and alluvial channels in forested mountain drainage basins. Nature 381, 587–589. Niedz´wiedz´, T., Obr˛ebska-Starklowa, B., 1991. Klimat. In: Dynowska, I. and Maciejewski, M. (Eds), Dorzecze go´rnej Wis"y. PWN, Warszawa-Krako´w, pp. 68–84. Osuch, B., 1968. Problemy wynikaja˛ ce z nadmiernej eksploatacji kruszywa rzecznego na przyk"adzie rzeki Wis"oki. Zesz. Nauk. Akademii Go´rniczo-Hutniczej 219, 283–301. Pie´gay, H., Darby, S.E., Mosselman, E., Surian, N., 2005. A review of techniques available for delimiting the erodible river corridor: a sustainable approach to managing bank erosion. River Res. Appl. 21, 773–789. Pietrzak, M., 2005. Relationship between settlement and relief in the Polish Carpathian mountains. In: Zg"obicki, W. and Rejman, J. (Eds), Human Impact on Sensitive Geosystems. Maria Curie Sk"odowska University Press, Lublin, pp. 65–82. Punzet, J., 1981. Zmiany w przebiegu stano´w wody w dorzeczu go´rnej Wis"y na przestrzeni 100 lat (1871–1970) (Changes in the course of water stages in the upper Vistula River basin over a century 1871–1970). Folia Geogr., Ser. Geogr. -Phys., Vol. 14, pp. 5–28 (in Polish, with English summary). Raczyn´ski, K., 1989. Wymiarowanie koryt cieko´w go´rskich (Dimensioning of mountain stream beds). Gosp. Wodna 4, 76–78 (in Polish, with English summary). Radecki-Pawlik, A., 2002. Pobo´r z˙ wiru i otoczako´w z dna potoko´w go´rskich (Mining of gravel and cobbles from mountain stream beds). Aura 2, 17–19 (in Polish, with English summary). Reid, I., Frostick, L.E., Layman, J.T., 1985. The incidence and nature of bedload transport during ﬂood ﬂows in coarse-grained alluvial channels. Earth Surf. Process. Landf. 10, 33–44. Richards, K., Clifford, N., 1991. Fluvial geomorphology: structured beds in gravelly rivers. Prog. Phys. Geogr. 15, 407–422.

A review on channel incision in the Polish Carpathian rivers

553

Rinaldi, M., 2003. Recent channel adjustments in alluvial rivers of Tuscany, central Italy. Earth Surf. Process. Landf. 28, 587–608. Rinaldi, M., Wyz˙ ga, B., Surian, N., 2005. Sediment mining in alluvial channels: physical effects and management perspectives. River Res. Appl. 21, 805–828. Schumm, S.A., 1969. River metamorphosis. J. Hydraul. Div. Am. Assoc. Civ. Eng. 95, 255–273. Sear, D.A., Archer, D., 1998. Effects of gravel extraction on stability of gravel-bed rivers: the Wooler Water, Northumberland, UK. In: Klingeman, P.C., Beschta, R.L., Komar, P.D., and Bradley, J.B. (Eds), GravelBed Rivers in the Environment. Water Resources Publications, Highlands Ranch, CO, pp. 415–432. Simon, A., 1989. A model of channel response in disturbed alluvial channels. Earth Surf. Process. Landf. 14, 11–26. Soja, R., 1977. Deepening of channel in the light of the cross proﬁle analysis (Carpathian river as example). Stud. Geomorph. Carp.-Balcan. 11, 127–138. Surian, N., Rinaldi, M., 2003. Morphological response to river engineering and management in alluvial channels in Italy. Geomorphology 50, 307–326. Szuman´ski, A., 1986. Postglacjalna ewolucja i mechanizm transformacji dna doliny dolnego Sanu (Late Glacial evolution and mechanism of transformation of the ﬂoor of the lower San valley). Geologia 12, 5–92. Williams, G.P. and Wolman, M.G., 1984. Downstream effects of dams on alluvial rivers. U. S. Geol. Surv. Prof. Paper, 1286, 1–83. Wyz˙ ga, B., 1991. Present-day downcutting of the Raba River channel (Western Carpathians, Poland) and its environmental effects. Catena 18, 551–566. Wyz˙ ga, B., 1993a. Present-day changes in the hydrologic regime of the Raba River (Carpathians, Poland) as inferred from facies pattern and channel geometry. In: Marzo, M. and Puigdefa´bregas C. (Eds), Alluvial Sedimentation. Intern. Assoc. Sediment. Spec. Publ., Vol. 17, pp. 305–316. Wyz˙ ga, B., 1993b. River response to channel regulation: case study of the Raba River, Carpathians, Poland. Earth Surf. Process. Landf. 18, 541–556. Wyz˙ ga, B., 1993c. Funkcjonowanie systemu rzecznego s´ rodkowej i dolnej Raby w ostatnich 200 latach (Evolution of the ﬂuvial system of the middle and lower Raba River (Carpathians, Poland) in the last 200 years). Dokument. Geogr. 6, 1–92 (in Polish, with English summary). Wyz˙ ga, B., 1996. Changes in the magnitude and transformation of ﬂood waves subsequent to the channelization of the Raba River, Polish Carpathians. Earth Surf. Process. Landf. 21, 749–763. Wyz˙ga, B., 1997. Methods for studying the response of ﬂood ﬂows to channel change. J. Hydrol. 198, 271–288. Wyz˙ ga, B., 1999. Estimating mean ﬂow velocity in channel and ﬂoodplain areas and its use for explaining the pattern of overbank deposition and ﬂoodplain retention. Geomorphology 28, 281–297. Wyz˙ ga, B., 2001a. A geomorphologist’s criticism of the engineering approach to channelization of gravelbed rivers: case study of the Raba River, Polish Carpathians. Environ. Manage. 28, 341–358. Wyz˙ ga, B., 2001b. Impact of the channelization-induced incision of the Skawa and Wis"oka Rivers, southern Poland, on the conditions of overbank deposition. Regul. Rivers Res. Manage. 17, 85–100. Zawiejska, J., Krzemien´, K., 2004. Human impact on the dynamics of the upper Dunajec River channel: a case study. Geograﬁcky Cˇasopis 56, 111–124. Zawiejska, J., Wyz˙ ga, B., 2005. Patterns, causes and controls of the 20th-century channel changes of the Dunajec River, southern Poland. Abstracts of 6th Gravel-Bed Rivers Conference, pp. 156–158.

Discussion by R.J. Batalla and A. Rovira Gravel mining appears to explain an important part of the incision phenomena observed in many European rivers throughout the second half of the 20th century, besides other human impacts, such as river embankment, dams and land-use change. However, there is still no general concept foreseeing the recovery in time and place of the river channel’s original equilibrium after mining ceases.

A review on channel incision in the Polish Carpathian rivers

553

Rinaldi, M., 2003. Recent channel adjustments in alluvial rivers of Tuscany, central Italy. Earth Surf. Process. Landf. 28, 587–608. Rinaldi, M., Wyz˙ ga, B., Surian, N., 2005. Sediment mining in alluvial channels: physical effects and management perspectives. River Res. Appl. 21, 805–828. Schumm, S.A., 1969. River metamorphosis. J. Hydraul. Div. Am. Assoc. Civ. Eng. 95, 255–273. Sear, D.A., Archer, D., 1998. Effects of gravel extraction on stability of gravel-bed rivers: the Wooler Water, Northumberland, UK. In: Klingeman, P.C., Beschta, R.L., Komar, P.D., and Bradley, J.B. (Eds), GravelBed Rivers in the Environment. Water Resources Publications, Highlands Ranch, CO, pp. 415–432. Simon, A., 1989. A model of channel response in disturbed alluvial channels. Earth Surf. Process. Landf. 14, 11–26. Soja, R., 1977. Deepening of channel in the light of the cross proﬁle analysis (Carpathian river as example). Stud. Geomorph. Carp.-Balcan. 11, 127–138. Surian, N., Rinaldi, M., 2003. Morphological response to river engineering and management in alluvial channels in Italy. Geomorphology 50, 307–326. Szuman´ski, A., 1986. Postglacjalna ewolucja i mechanizm transformacji dna doliny dolnego Sanu (Late Glacial evolution and mechanism of transformation of the ﬂoor of the lower San valley). Geologia 12, 5–92. Williams, G.P. and Wolman, M.G., 1984. Downstream effects of dams on alluvial rivers. U. S. Geol. Surv. Prof. Paper, 1286, 1–83. Wyz˙ ga, B., 1991. Present-day downcutting of the Raba River channel (Western Carpathians, Poland) and its environmental effects. Catena 18, 551–566. Wyz˙ ga, B., 1993a. Present-day changes in the hydrologic regime of the Raba River (Carpathians, Poland) as inferred from facies pattern and channel geometry. In: Marzo, M. and Puigdefa´bregas C. (Eds), Alluvial Sedimentation. Intern. Assoc. Sediment. Spec. Publ., Vol. 17, pp. 305–316. Wyz˙ ga, B., 1993b. River response to channel regulation: case study of the Raba River, Carpathians, Poland. Earth Surf. Process. Landf. 18, 541–556. Wyz˙ ga, B., 1993c. Funkcjonowanie systemu rzecznego s´ rodkowej i dolnej Raby w ostatnich 200 latach (Evolution of the ﬂuvial system of the middle and lower Raba River (Carpathians, Poland) in the last 200 years). Dokument. Geogr. 6, 1–92 (in Polish, with English summary). Wyz˙ ga, B., 1996. Changes in the magnitude and transformation of ﬂood waves subsequent to the channelization of the Raba River, Polish Carpathians. Earth Surf. Process. Landf. 21, 749–763. Wyz˙ga, B., 1997. Methods for studying the response of ﬂood ﬂows to channel change. J. Hydrol. 198, 271–288. Wyz˙ ga, B., 1999. Estimating mean ﬂow velocity in channel and ﬂoodplain areas and its use for explaining the pattern of overbank deposition and ﬂoodplain retention. Geomorphology 28, 281–297. Wyz˙ ga, B., 2001a. A geomorphologist’s criticism of the engineering approach to channelization of gravelbed rivers: case study of the Raba River, Polish Carpathians. Environ. Manage. 28, 341–358. Wyz˙ ga, B., 2001b. Impact of the channelization-induced incision of the Skawa and Wis"oka Rivers, southern Poland, on the conditions of overbank deposition. Regul. Rivers Res. Manage. 17, 85–100. Zawiejska, J., Krzemien´, K., 2004. Human impact on the dynamics of the upper Dunajec River channel: a case study. Geograﬁcky Cˇasopis 56, 111–124. Zawiejska, J., Wyz˙ ga, B., 2005. Patterns, causes and controls of the 20th-century channel changes of the Dunajec River, southern Poland. Abstracts of 6th Gravel-Bed Rivers Conference, pp. 156–158.

Discussion by R.J. Batalla and A. Rovira Gravel mining appears to explain an important part of the incision phenomena observed in many European rivers throughout the second half of the 20th century, besides other human impacts, such as river embankment, dams and land-use change. However, there is still no general concept foreseeing the recovery in time and place of the river channel’s original equilibrium after mining ceases.

554

B. Wyz˙ga

The Tordera River (NE Spain) is a representative case of water and sediment dynamics in a large river in the Western Mediterranean region, which was intensively mined for aggregate between the 1960s and 1980s. Approximately 3 106 m3 of gravel were extracted from the river’s mainstem. Mining itself caused mean channel incision of 1.5 m (reaching 4 m in several sections), undermining bridges and overdrafting the groundwater in an area with intense demand, especially for agricultural and tourist uses. The sand supply to the beaches along the coast has also been affected. Channel margins were stabilized by rip-rap in most river reaches. Field data (Rovira et al., 2005) indicates that 15 years after mining ceased, the river shows during dry periods a general tendency towards aggradation (3.6 mm/yr on average and up to 12 mm/yr in central sections where mining was more intensive) but a slight long-term tendency to erode the riverbed, especially in the lowermost sections where the channel geometry was not modiﬁed. The potential of the river to increase its channel elevation in the future is limited to the dry periods whereas in the long term, slight degradation of the riverbed is expected until the river reaches a new dynamic equilibrium. The persistence of huge gravel pits in the long proﬁle that today trap the coarser fractions, reinforces the argument that the aggradation is probably not representative of a morphological trend over a long time period. Overall, the estimated time for the river to recover the pre-extraction bed-level would be around 420 years. These observations may provide insights into the variable behaviour of the river channel response (from persistent channel incision to almost river-bed level stabilization) observed in several rivers of the Polish Carpathians after mining operations ended several years ago. Author may consider the role played by the relative position of the control sections in the channel network in relation to the areas of gravel extraction.

Reference Rovira, A., Batalla, R.J., Sala, M., 2005. Response of a river sediment budget after historical gravel mining (the lower Tordera, NE Spain). River Res. Appl. 21, 829–847.

Reply by the author The case of the Tordera from NE Spain and those of the Ropa and Wis"oka described in the paper illustrate the substantial impact exerted on river channels by the in-stream sediment mining conducted at a rate greatly exceeding the rate of material replenishment from upstream. Following the large-scale gravel mining carried out in the two Carpathian rivers between the 1940s and 1960s, the Ropa channel has undergone transformation from alluvial to bedrock boundary conditions over the length of a few tens of kilometres whereas the reach of the Wis"oka channel with the most intensive exploitation incised by about 2.5 m during the second half of the 20th century. Apart from the high rate of material extraction, a few other factors have made the impact of the sediment mining on the Carpathian rivers especially

554

B. Wyz˙ga

The Tordera River (NE Spain) is a representative case of water and sediment dynamics in a large river in the Western Mediterranean region, which was intensively mined for aggregate between the 1960s and 1980s. Approximately 3 106 m3 of gravel were extracted from the river’s mainstem. Mining itself caused mean channel incision of 1.5 m (reaching 4 m in several sections), undermining bridges and overdrafting the groundwater in an area with intense demand, especially for agricultural and tourist uses. The sand supply to the beaches along the coast has also been affected. Channel margins were stabilized by rip-rap in most river reaches. Field data (Rovira et al., 2005) indicates that 15 years after mining ceased, the river shows during dry periods a general tendency towards aggradation (3.6 mm/yr on average and up to 12 mm/yr in central sections where mining was more intensive) but a slight long-term tendency to erode the riverbed, especially in the lowermost sections where the channel geometry was not modiﬁed. The potential of the river to increase its channel elevation in the future is limited to the dry periods whereas in the long term, slight degradation of the riverbed is expected until the river reaches a new dynamic equilibrium. The persistence of huge gravel pits in the long proﬁle that today trap the coarser fractions, reinforces the argument that the aggradation is probably not representative of a morphological trend over a long time period. Overall, the estimated time for the river to recover the pre-extraction bed-level would be around 420 years. These observations may provide insights into the variable behaviour of the river channel response (from persistent channel incision to almost river-bed level stabilization) observed in several rivers of the Polish Carpathians after mining operations ended several years ago. Author may consider the role played by the relative position of the control sections in the channel network in relation to the areas of gravel extraction.

Reference Rovira, A., Batalla, R.J., Sala, M., 2005. Response of a river sediment budget after historical gravel mining (the lower Tordera, NE Spain). River Res. Appl. 21, 829–847.

Reply by the author The case of the Tordera from NE Spain and those of the Ropa and Wis"oka described in the paper illustrate the substantial impact exerted on river channels by the in-stream sediment mining conducted at a rate greatly exceeding the rate of material replenishment from upstream. Following the large-scale gravel mining carried out in the two Carpathian rivers between the 1940s and 1960s, the Ropa channel has undergone transformation from alluvial to bedrock boundary conditions over the length of a few tens of kilometres whereas the reach of the Wis"oka channel with the most intensive exploitation incised by about 2.5 m during the second half of the 20th century. Apart from the high rate of material extraction, a few other factors have made the impact of the sediment mining on the Carpathian rivers especially

A review on channel incision in the Polish Carpathian rivers

555

severe and difﬁcult to reverse in the future (cf. Rinaldi et al., 2005): (i) location of the exploitation in channelized reaches where bank-protection structures prevent a delivery of bed material from eroded channel banks and the adjustment of transport capacity of the watercourses through an increase in channel sinuosity; (ii) the concurrent reafforestation of the montane part of the catchments, reducing the upstream sediment supply (Lach and Wyz˙ ga, 2002); and (iii) a thin cover of alluvium underlying both rivers, especially the Ropa, which has facilitated exhaustion of gravel in the channels, hence making the subsequent bed recovery a very slow process. Indeed, no signiﬁcant recovery of the Ropa River bed has occurred following incision of its channel to bedrock. This is indicated by the relatively stable vertical position of minimum annual water stages recorded during the two last decades at the Kl˛eczany and Topoliny gauging stations, upstream and downstream of the mining site, respectively (see Fig. 4 in Rinaldi et al., 2005). At the Brzez´nica and Łabuzie stations located in the reach of the Wis"oka River with the most intensive gravel extraction, channel incision continued during the two decades after the cessation of the in-stream mining in the late 1960s. Stabilization of the bed level in the 1990s (Fig. 3) indicates that the river might have then attained a new dynamic equilibrium. At the Krajowice station located about 30 km upstream of this reach, bedrock control prevented any channel incision during the second half of the 20th century. At Mielec, 20 km downstream of the reach, the river bed remained vertically stable during the period of the upstream-conducted sediment mining but incised rapidly by more than 1 m during the 1970s–1980s (Fig. 5) in response to channelization works and, probably, also the upstream-borne sediment deﬁcit. While the spacing of the gauging stations on the river is too large to determine exactly the longitudinal extent of channel incision induced by the in-stream mining, undoubtedly no increase in bed elevation occurred during the 40 years following the cessation of the mining activity.

References Lach, J., Wyz˙ ga, B., 2002. Channel incision and ﬂow increase of the upper Wis"oka River, southern Poland, subsequent to the reafforestation of its catchment. Earth Surf. Process. Landf. 27, 445–462. Rinaldi, M., Wyz˙ ga, B., Surian, N., 2005. Sediment mining in alluvial channels: physical effects and management perspectives. River Res. Appl. 21, 805–828.

Gravel-Bed Rivers VI: From Process Understanding to River Restoration H. Habersack, H. Pie´gay, M. Rinaldi, Editors r 2008 Elsevier B.V. All rights reserved.

557

21 Contemporary morphological change in braided gravel-bed rivers: new developments from ﬁeld and laboratory studies, with particular reference to the inﬂuence of riparian vegetation D. Murray Hicks, Maurice J. Duncan, Stuart N. Lane, Michal Tal and Richard Westaway

Abstract Contemporary (event to decadal-scale) morphological changes in two large braided rivers in Canterbury, New Zealand, are described, along with laboratory studies that support the ﬁeld observations. In the process, some new developments in ﬁeld and laboratory methods for investigating morphological change in braided rivers are presented, and Paola’s (2001) hypothesis that braiding tendency should be inﬂuenced by a river’s ability to turn over its bed within the characteristic time for riparian vegetation to establish and grow to a mature, scour-resistant state is examined. The lower Waitaki River has been regulated for hydropower since 1935, and since then vegetation has encroached over the riverbed and braiding intensity for a given discharge has reduced. Measurements of vegetation removal by ﬂoods indicate that ﬂoods are not able to turn over the bed fast enough to contain vegetation encroachment, and the present braiding state is held by virtue of a regular spraying programme. In contrast, on the unregulated and sparsely vegetated Waimakariri River, remotely sensed high-resolution topographic surveys using LiDAR and digital photogrammetry have shown that even sub-annual ﬂoods turn over large proportions of the braidplain. The laboratory studies show that a braided river will evolve into a single-thread channel when its bed is invaded by vegetation and ﬂoods are too infrequent to contain the vegetation growth. Collectively, the ﬁeld and laboratory evidence conﬁrms that Paola’s (2001) dimensionless time-scale parameter is a reasonable ﬁrst-order predictor of whether ﬂoods or vegetation will achieve ascendancy, driving a river towards either braided or single-thread end-points, respectively.

E-mail address: [email protected] (D.M. Hicks) ISSN: 0928-2025

DOI: 10.1016/S0928-2025(07)11143-3

D.M. Hicks et al.

558 1.

Introduction

Braided gravel-bed rivers are spectacular and relatively unconstrained features of the ﬂuvial landscape. They offer a special mixture of physical habitats that underpin diverse and unique ecosystems (e.g., Glova and Duncan, 1985; Maloney et al., 1997; Kilroy et al., 2004). Braiding arises from a fuzzy recipe of high and usually frequently applied stream power (i.e., slope and ﬂood discharges), relatively cohesionless and erodible banks composed of bedload-grade sediment, and an abundant external supply of this bedload material (Knighton and Nanson, 1993; Knighton, 1998). These controls combine to deliver morphologies that are as much characterised by their dynamic behaviour as by their synoptic appearance (Ashmore, 2001). Human impacts on these drivers, usually through dams and/or water abstractions, typically damp ﬂood regimes and alter relative sediment supplies (Young et al., 2004). Changes in river morphology triggered by these impacts may take decades to develop. This time scale matches the term of many resource consent applications and licences, yet most current physical habitat assessment methodologies, and indeed consent decisions, assume that river channel morphology is ﬁxed – that it will not change over the consent period. Thus a primary motivation for studying morphological change in these systems is to be able to predict how instream habitat might evolve following human-induced changes in ﬂow and sediment transport regime over decadal time scales. Another motivation is to study the short-term, ﬂood-by-ﬂood patterns of erosion and deposition to gain insight into braiding processes, estimate bedload transport by morphological methods (e.g., Lane et al., 1995), and assess the stability of physical habitat. Before any of these can be done, however, it is ﬁrst necessary to quantify the change that has occurred. Measuring the morphology of large braided river systems and monitoring contemporary changes have until recently been conﬁned to two dimensions, by way of vertical aerial imagery or cross-section surveys (e.g., Goff and Ashmore, 1994). Moreover, the aerial perspective has focussed on features of wetted areas rather than morphology per se. Now, differential GPS, laser-based total station instruments, and ground-based laser scanning provide the opportunity to gather full topography of channels at scales of 1 to several 100 m (e.g., Brasington et al., 2000; Fuller et al., 2003; Heritage and Hetherington, 2007). Also, and of prime value in braided river environments, aerial remote-sensing approaches (e.g., digital photogrammetry and LiDAR) have enabled fully 3D topography to be gathered on a synoptic basis over large areas many km in dimension (e.g., Lane et al., 2003). LiDAR, in particular, has emerged as a potent tool because of its ability to ‘‘see through’’ vegetation while also capturing information useful for classifying ground cover. In addition, LiDAR systems able to map through water in shallow ﬂuvial channels have recently come available (Millar et al., 2005). Another key requirement for predicting morphological change is a better understanding of the effect of riparian vegetation. Riparian vegetation is a fundamental control on river channel morphology (Simon et al., 2004; Tal et al., 2004), and human-driven ﬂow regulation usually results in increased tendencies for riverbed colonisation by vegetation (Johnson, 1997). The additional drag and scour resistance provided by mainly woody vegetation on banks and ﬂoodplains, and feedback-loops

Contemporary morphological change in braided gravel-bed rivers

559

that encourage aggradation on vegetated areas, are key factors in the transition from truly braiding systems (i.e., where the braidplain is inundated during ﬂood ﬂows and becomes, effectively, an active ﬁeld of low-relief bedforms) towards anastomosing systems where vegetated islands persist (Knighton and Nanson, 1993). Recently, vegetation-inﬂuenced bank strength has been incorporated into a function that discriminates between braiding and meandering (Millar, 2000). Vegetation can be very difﬁcult to see and walk through, however, and so, in the ﬁeld at least, its impacts on morphology have not generally been well studied. The new technologies discussed above help ease this problem. A common ﬁeld problem is that several controls can change simultaneously so that it is typically difﬁcult to isolate the effects of one control (e.g., dams can reduce sediment supply and regulate ﬂows, which can both facilitate encroachment of vegetation). In this regard, laboratory and numerical models – where all other factors can be held ﬁxed – are useful for inspiring questions, testing hypotheses, identifying processes and key parameters, accelerating time scales, and quantifying relationships. In particular, some recent physical and numerical modelling work in these areas has advanced our understanding of the importance of riparian vegetation (e.g., Gran and Paola, 2001; Murray and Paola, 2003; Coulthard, 2005; Tal and Paola, 2007). Also, numerical modelling is now used to examine river morphological change at basin scales (Coulthard et al., 2007; Pizzuto et al., 2007). In this paper our aim is to present some new developments in ﬁeld methods and laboratory studies for investigating contemporary morphological change in braided rivers, with an emphasis on applying these to assessing the inﬂuence of invasive riparian vegetation on driving morphologic change. We illustrate these with case examples from two large braided rivers in Canterbury, New Zealand: one (the Waitaki River) already undergoing change from human activities that have altered its ﬂow and sediment regime and where the braidplain has been invaded by exotic vegetation, the other (the Waimakariri River) with a largely natural ﬂow and sediment regime and un-vegetated braidplain. We begin with a brief conceptual review of how riparian vegetation may combine with other drivers to effect morphological change. Next we present the case example of the Waitaki River to illustrate typical changes in key drivers and what the morphological response has been. We describe new methods for and results from measuring fully 3D morphological change as developed for the Waimakariri River. We then describe some current laboratory work that examines the processes and morphologic consequences when a braided riverbed is vegetated. We end by discussing the play-off between riparian vegetation and ﬂoods, the ongoing need for measurements of morphologic change, and offer an opinion on the best way forward for measuring morphological change.

2.

Riparian vegetation effects on channel processes and properties: a brief review

As reviewed in Tal et al. (2004), riparian vegetation impacts a variety of ﬂow and sedimentation processes and bed/bank properties. At local scales, vegetation can inﬂuence ﬂow velocities and depths by imposing additional drag and by physically reducing the space through which water ﬂows (Nepf, 1999; Jarvela, 2004; Green,

560

D.M. Hicks et al.

2005). Velocity reduction within plant communities hinders scour and indeed encourages the deposition of suspended sediment, while velocity reduction in the wakes of copses and large trees can focus bed-material accumulation to form proto-islands (e.g., Coulthard, 2005). In turn, deposition encourages further vegetation growth, which permits bars and islands to increase in height and attain greater permanency. Vegetation also imposes strength or ‘cohesion’ to non-cohesive bed and bank material through several processes and effects, including protecting underlying substrate by litter and leaf mats and protecting banks by large-woody debris, root strength, and water-capillary effects (Simon and Collinson, 2002). At channel scales, through the above processes, riparian vegetation encourages channel narrowing by both hindering bank erosion and assisting bank accretion (Thorne, 1990; Lie´bault and Pie´gay, 2002). Moreover, both the additional drag imposed by riparian vegetation and the rising elevation of islands and banks through sedimentation serve to concentrate ﬂood ﬂows into low-ﬂow channels. These effects can drive a positive feedback loop involving channel incision, which further increases the relative elevation between channel beds and bar and bank tops, leading to additional ﬂow capture and less frequent over-bank ﬂooding (Tsujimoto and Kitamura, 1996, 1998). In turn, this further encourages the permanency of bank and bar vegetation, the positional stability of channels, and channel narrowing. Riparian vegetation cover is regulated by several time scales that are set by the seasons, the times for germination and growth-to-maturity of individual species, and the succession rates of species as riparian communities evolve. Johnson (1997, 2000) showed that vegetation recruitment (i.e., deposition and germination of seedlings) and survivability are inﬂuenced by water discharge, sedimentation, and the seasonal phasing of ﬂoods and plant reproduction. For example, for the Platte River (central Nebraska, USA), he showed that while willows and poplars produce large numbers of seeds in early summer, thus creating the potential for woodland expansion on the riverbed, the summer ﬂoods hindered recruitment by scouring seedlings. Johnson also showed that maturing riparian woodland becomes less effective at damping ﬂood velocities because it is more open. In a sense, at least in temperate climates, riverbeds are arenas in which vegetation attempts to establish itself while the rivers try to scour it away. In consequence, going up the banks, a succession of vegetation is encountered that has adapted to a progressively reducing frequency of inundation and which reﬂects the prevailing hydrological regime (Hupp and Osterkamp, 1996). Braided rivers represent a special case where ﬂoodplain exposure to a given ﬂood discharge occurs on a probabilistic basis rather than a deterministic one. Braided rivers are typiﬁed by wide active channels in non-cohesive sediment that migrate laterally and, in the process, both consume and generate bars (that become islands at low ﬂow). Inevitably, the active channels tend to favour only part of the valley ﬂoor at any one time, and in their absence, areas of ﬂoodplain form in patches as bars are capped with vertically accreting ﬁne sediment (Nanson and Croke, 1992; Reinfelds and Nanson, 1993; Knighton, 1998). Vegetation aids this accretionary process by trapping and binding ﬁne sediment. Thus, the time between ﬂood inundations of any part of the riverbed varies on a joint probabilistic basis associated with ﬂood frequency, location of main channels, and to some extent the past history of the riverbed. While the probability of

Contemporary morphological change in braided gravel-bed rivers

561

inundation reduces away from the valley centre so that vegetation is more likely to grow into the riverbed from the valley margins, it is not precluded from growing on islands within the riverbed as well. A face-off must occur on riverbeds between the space/time characteristics of seed dispersal and plant growth and the occupation, abandonment, and reworking of the bed by ﬂowing water. Vegetation that is not removed while young and weakly rooted will become increasingly resistant to scour. Thus a key organizing parameter must be the time scale for establishment of the vegetation relative to a characteristic channel or bed mobility time. Paola (2001) developed the idea that the continuum of meandering-braided channels should reﬂect the ratio of the time required for vegetation to grow to a sufﬁciently mature state that it can resist scour (Tveg) compared with the average interval between scour events. This can be indexed by a dimensionless time scale T* ¼ TvegE/B, where E is the average horizontal rate at which the riverbed is eroded by channel migration or scour and B is the width of active channel bed, hence B/E is the time required for the channel bed to be completely reworked and cleared of vegetation. B/E therefore becomes a function of the magnitude–frequency distribution of ﬂoods. At one end of the time-scale continuum, vegetation can never become established due to frequent ﬂooding and very dynamic migration of channels, in which case a braided system is maintained. At the other end of the continuum, vegetation permanently colonizes all but the minimum width required for a single (meandering) channel to transport the sediment load. Thus vegetation-driven (or at least catalysed) changes in braided channel morphology are likely to stem from a damping of the work done along channel margins, the progressive isolation of bar and island surfaces from competent ﬂows, the polarisation of bedload transport processes into fewer, more permanent channels, and any down-shift in the frequency of scour-competent ﬂoods relative to the time-scale of vegetation establishment. Moreover, due to the probabilistic nature of both ﬂoods and channel response, changes will develop over an extended period and may be difﬁcult to identify against natural variability.

3.

The lower Waitaki River: changed drivers and morphologic response

The Waitaki River (Fig. 21.1) is New Zealand’s largest braided river by discharge (mean discharge E370 m3/s) and a major source of hydroelectric power. Hydropower works began in 1935 and include three dams along the middle, gorged section of the Waitaki Valley and a network of canals, control structures, and power stations that utilize the storage from three large natural lakes in the upper basin. The 70 km long reach of river between the last dam and the coast is termed the lower Waitaki. It is braided for all but the ﬁrst 5 km downstream of the dam, and has an average valley slope of 0.003. The lower Waitaki is a popular ﬁshing river for trout, salmon, and indigenous species, and is equally popular for jet boating. Eighty percent of the ﬂow of the lower Waitaki River derives from the Southern Alps, along the north-western catchment boundary, and passes through the three natural lakes. Storage in these lakes damps the lower Waitaki’s ﬂood regime relative to those of the other braided rivers that drain the Southern Alps (such as the

562

D.M. Hicks et al.

Figure 21.1. The Waitaki and Waimakariri Rivers, South Island, New Zealand.

Waimakariri), since these have less or no signiﬁcant lake storage in their catchments (Figs. 21.2 and 21.3a). The natural runoff is seasonal, with a winter low and summer high. Most ﬂoods stem from spring–summer rainfall or rain-plus-snowmelt events in the Southern Alps. The upstream hydro-works have further damped the ﬂow regime, with reduced ﬂood magnitudes (the mean annual ﬂood has reduced from 1350 to 1110 m3/s, based on recorded and simulated natural daily mean ﬂows over a reference period), generally steadier ﬂows, and less seasonality (Fig. 21.3). The hydro dams have reduced the supply of bed-material to the lower Waitaki by approximately 50% (Hicks et al., 2003a). The remaining bed-material is sourced from tributaries and reworking of the Pleistocene valley-ﬁll. The bed material is dominated by greywacke gravel, with a subsidiary fraction (18%) of ﬁne-medium sand. The overall median size of the subsurface bed material (averaged from 28 approximately 200 L samples dug from bars along the lower river) is 22 mm, while the median size of the gravel fraction is 32 mm. The maximum clast size (intermediate axis) observed anywhere along the lower Waitaki is 370 mm. When surveyed in 2005, the bed surface was characterized by cobble-armoured channels and scoured bars, with scattered sheets and lobes of ﬁner, mobile gravels. The median armour size measured in channels and on bars ranged from 42 to 97 mm

Contemporary morphological change in braided gravel-bed rivers

563

3000 Flow m3/s

Waitaki at Waitaki Dam 2000

0 19900101

9301

9601

9901

0201 YYMM

9601

9901

0201 YYMM

3000 Flow m3/s

Waimakariri at SH1 Bridge 2000

0 19900101

9301

30

Monthly mean discharge / mean discharge

Peak discharge / mean discharge

Figure 21.2. Hydrographs of the lower Waitaki and Waimakariri Rivers for the period 1990–2004.

Waitaki natural (simulated) 25

Waitaki actual Waimakariri

20 15 10

(a)

5 0

1

10 Return period (yr)

100

(b)

1.6 1.4 1.2 1 0.8 0.6 0.4 0.2

Waitaki natural (simulated) Waitaki actual Waimakariri

0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Month

Figure 21.3. (a) Flood magnitude vs. return period (partial duration series) for the Waitaki River at Waitaki Dam actual ﬂow record, a simulated natural record for the same location, and the Waimakariri River at State Highway 1 Bridge for the period 1967–2004. (b) Monthly mean discharge at the same sites over the same period. Discharges are normalized by mean discharge.

and averaged 70 mm. The median surface size of the gravel sheets clustered closely around 30 mm. Sand was scarce on the bed surface except in low-energy locations such as backwaters. Vegetated islands typically had a thin capping of silt deposited over an often-armoured surface. No downstream ﬁning trend was apparent in either the subsurface or surface material along the 65 km braided reach. Bedload transport begins when the river discharge exceeds 450 m3/s but becomes more substantial when the ﬂow exceeds 900 m3/s, which approximates the discharge required to fully inundate the relatively un-vegetated fairway (based on both casual observation and calculation). Bedload movement is typically manifest by the migration of the sparsely spaced gravel sheets and lobes. These appear to be initiated

564

D.M. Hicks et al.

mainly through phases of bank erosion, and they migrate a short distance downstream before stalling. Flow events when the discharge exceeds 900 m3/s occur approximately once per two years on average and typically span several days but may last several weeks.

3.1.

Morphological response to changed flow and sediment regimes

While some of the sediment supply deﬁcit induced by the Waitaki hydro dams has been recovered by degradation within a few km downstream of Waitaki Dam, degradation along the braided reach is not obvious; indeed, estimates of bedload transport capacity using the surface-based transport model of Wilcock and Crowe (2003) show that the reduced sediment supply has been balanced by reduced transport capacity of the ﬂow regime (so that the relative supply has stayed about the same). Thus, spatially extensive armouring similar to that observed in recent years, which reﬂects a low gravel supply relative to bedload transport capacity, was likely also a feature of the riverbed in its natural state. The pre-dam riverbed of the lower Waitaki was up to 2 km wide and characterized by sparse willow trees, temporary islands vegetated mainly by native tussock and scrub, and shifting gravel bars and channels (Fig. 21.4a). The onset of ﬂow regulation was followed by an invasion of the riverbed by willow, broom, and gorse. These exotic species were introduced by Europeans in the late 19th and early 20th centuries, and appeared able to establish in the lower Waitaki post ﬂow control because the regulated ﬂow regime lacked the extreme seasonal variation in ﬂows that naturally ﬂushed seedlings from the bed and allowed stock access to and grazing of the islands during low winter ﬂows. In consequence, the less resilient native vegetation was displaced and islands and bars became choked with the exotic vegetation and tended to stabilize, while ﬂood breakouts along the riverbed margins became a hazard (Hall, 1984). Although a policy of de-vegetating a central ‘fairway’ with spraying and machinery has been implemented since the 1960s, a time-series of aerial photographs shows a trend of increasing vegetation cover and reducing fairway width that has continued to the present. By 2001, the un-vegetated part of the river’s braidplain had been reduced to an average width of about 0.5 km. Studies of aerial photographs (e.g., Hall, 1984; Tal et al., 2004) have noted that this narrowing of the braidplain has been accompanied by a reduction in braiding activity and a tendency for ﬂows to congregate in one or two principle braids (Fig. 21.4b). For the reach shown in Fig. 21.4, the average number of braids at 150 m3/s decreased from 12 to 7. Two-dimensional hydrodynamic modelling using the 2001 ground-cover and also the pre-control ground-cover (Hicks et al., 2006) showed that the additional drag developed by the recent vegetation has reduced ﬂood velocities over the vegetated islands and fairway margins, stabilizing their beds and encouraging suspended sediment deposition while focussing ﬂows into and increasing the bed shear stress in the main channels. Over time, positive feedback from these processes has encouraged the main channels to entrench relative to the island and bar levels, promoting the observed trend for positional stability.

Contemporary morphological change in braided gravel-bed rivers

565

Figure 21.4. Identical coverage of a 3.7 km span of the lower Waitaki River photographed in 1936 (a) and 2001 (b) when the river discharge was 150 m3/s. River ﬂows left to right. From Tal et al. (2004).

3.2.

Effectiveness of floods at clearing vegetation

We were able to assess the effectiveness of different sized ﬂoods at clearing riverbed vegetation (and, by implication, inducing morphological change) by examining the changes in riverbed ground cover and channel position on sequential airphotographs taken in 1994, 1995, and 1996. These captured the changes due to a 5-year return period ﬂood, peaking at 1488 m3/s in November 1994, and a near 200year ﬂood, peaking at approximately 3000 m3/s in December 1995. The photography was rectiﬁed to a common projection, and the ground cover mapped before and after each ﬂood across 117 valley-normal transects spaced 500 m apart along the whole 65 km length of river. Also, ground cover was mapped continuously over two 4-km long sub-reaches, with areas converted to average widths by dividing by the reach

D.M. Hicks et al.

566

Width of bare gravel & water (m)

length. Ground cover over the undeveloped braidplain was classiﬁed as trees (mainly willows), scrub (mainly gorse and broom), sparse scrub, grass, bare gravel, or wetted channel. Changes in the combined width of wetted channel and bare gravel along the 65 km river length illustrate the extent of vegetation clearance by the intervening ﬂoods (Fig. 21.5a). We estimate that the classiﬁcation procedure could have induced random errors of up to about 30 m in the total un-vegetated width at individual transects, but the uncertainty on the average change in width over all 117 transects reduces to approximately 6 m. Errors due to scaling differences associated with the photo-rectiﬁcation procedure amount to no more than approximately 1% of the measured widths. When averaged for the full 65 km length of river, the 5-year ﬂood cleared an average width of 27 m of vegetated riverbed, of which approximately half was low but dense scrubby vegetation (gorse and broom) and the rest mainly sparse scrub (Fig. 21.5b). This clearance equated to a 7% increase in the area of bare gravel and wetted channel. No signiﬁcant changes in the average width of trees (willows) were measured. In contrast, the 200-year ﬂood cleared an average 124 m width of vegetated riverbed, of which approximately 10% was willow, 59% was low scrub, and 28% was sparse scrub (Fig. 21.5b). This amounted to a 24% increase in the preﬂood area of bare gravel and wetted channel. The detailed mapping of the two subreaches showed that lateral channel shifts consumed approximately 1 m of tree and scrub-covered island or bank for every 2 m of bare gravel bar consumed, thus the a

1200 1000

1996

800

1994

1995

600 400 200 0 1

11

21

31 41 51 61 71 81 91 Transect (number increases downstream)

101

111

Average width (m)

1600

b

1200

Bare gravel & water Sparse scrub

800

Grass Scrub

400

Trees

0 1994 1995 1996 Year of photography Figure 21.5. Ground cover along Lower Waitaki River bed between Kurow and the coast, as captured from aerial photographs taken February 1994, February 1995, and March 1996. (a) Width of bare gravel and wetted channel at individual transects (transect number increases downstream). (b) Widths of all ground cover classes averaged over all 117 transects.

Contemporary morphological change in braided gravel-bed rivers

567

extent of bed ‘turnover’ during the ﬂoods was greater than was indicated simply by the reductions in vegetation cover as in Fig. 21.5. Based on these measurements, we conclude that (i) both ﬂoods scoured signiﬁcant areas of riverbed vegetation and induced shifts in the location of the main braids, (ii) the 200-year ﬂood was 4–5 times more effective at removing vegetation than was the 5-year ﬂood, but even the 200-year event only increased the area of un-vegetated riverbed by about one quarter, (iii) vegetated areas were less prone to ﬂood scour. The pragmatic ‘rule-of-thumb’ followed by river managers on the Waitaki is that a 5-yearly spraying programme is required to prevent the establishment of resilient woody vegetation within the riverbed. Assuming (i) that this represents the Tveg time scale introduced in Section 2, (ii) that the total average width of undeveloped braidplain is 1400 m (Fig. 21.5b), (iii) a lateral erosion rate of 81 m/yr (based on the measured consumption of 27 m of vegetated riverbed over the year containing the 5-year return-period ﬂood, scaled up by a factor of three to include lateral erosion of un-vegetated riverbed), then the dimensionless time scale, T*, E0.3. The average annual lateral erosion rate is most likely less than this assumed value, thus 0.3 is likely an upper bound for T* in the Waitaki. This value of T*{1 indicates that ﬂoods – on their own – should not be able to sweep the whole width of riverbed clear of the invasive exotic vegetation before it establishes in a scour-resistant, woody form. This conﬁrms, ﬁrst, the utility of Paola’s (2001) time-scale ratio and, second, the critical importance of the artiﬁcial vegetation control programme to maintaining the braided character of the lower Waitaki.

4. 4.1.

Measuring and mapping 3D morphological change in the Waimakariri River New surveying technologies

Technological advances over the past decade have made it possible to secure high resolution synoptic surveys of the 3D morphology of large areas of natural riverbed that rival those obtainable in the laboratory. Two independent advances have been digital photogrammetry and aerial LiDAR (Light Detection And Ranging, sometimes referred to as Aerial Laser Scanning or ALS). Example applications to braided rivers include Stojic et al. (1998), Lane (2000), Westaway et al. (2001, 2003), Charlton et al. (2003) and Chandler et al. (2004). Both technologies provide the ability to survey large areas (square km scale) of dry riverbed at high spatial density (point-spacings of the order of 1–2 m) with a similar vertical accuracy (0.1–0.2 m). While neither approach, at least until very recently, permits the mapping of submerged riverbed topography, supplementary techniques have been developed to achieve this. These include: ground/boat-based surveys of channel beds; calibrating colour aerial imagery or multi-spectral scans against water depth in order to map water depth (e.g., Winterbottom and Gilvear, 1997), which can then be differenced from water surface elevation obtained by LiDAR or photogrammetry (e.g., Westaway et al., 2003; Hicks et al., 2006); and clear-water photogrammetry (Westaway et al., 2001). The remote-sensing approaches for surveying submerged

D.M. Hicks et al.

568

topography inevitably involve degraded accuracy, thus surveys are typically done at low river ﬂows to maximise the exposure of dry riverbed.

4.2.

Application to Waimakariri River

We have employed both digital photogrammetry and LiDAR to survey changing morphology on a study reach of the braided Waimakariri River on the Canterbury Plains, New Zealand. The motivation for this was threefold: (i) to obtain digital elevation models (DEMs) for use with ﬁxed-bed 2D hydrodynamic models for physical habitat mapping; (ii) to survey erosion and deposition before and after ﬂood events in order to estimate bedload transport by the ‘morphological approach’; and (iii) to assess the extent of riverbed turnover due to ﬂoods. The latter question is pertinent both to the longevity of instream habitat and to the survival of riverbed vegetation. Here, we brieﬂy describe the Waimakariri study site, the approaches followed, the uncertainties in the measurements, and present example results relating to the extent of ﬂood-driven riverbed ‘turnover’ within the context of Paola’s (2001) dimensionless time scale, T*. The application of the results to erosion and deposition patterns and to bedload transport estimation are reported elsewhere (Hicks et al., 2002, 2003b; Lane et al., 2003). The 3-km long study reach on the Waimakariri River (Fig. 21.6) is located at Crossbank, 7 km north of the city of Christchurch on New Zealand’s South Island (Fig. 21.1). Along the study reach the river is conﬁned to a straight 1500-m wide fairway by stopbanks. Belts of willows inside the stopbanks, planted by the local river management authority, reduce the width of active riverbed to approximately 1000 m. By and large, this active area is naturally kept clear of vegetation by frequent freshes and ﬂoods, but patches of lupin typically sprout annually on the higher bars and islands during summer while isolated patches of broom and tussock may persist for several years on elevated areas remote from the main braids. Naturally, this reach was constructing an alluvial fan but following conﬁnement the fan deposition has

Figure 21.6. The Waimakariri River at Crossbank, photographed in February 2000 when the discharge was 65 m3/s. Flow is left to right.

Contemporary morphological change in braided gravel-bed rivers

569

been concentrated into a narrower width. The consequent aggradation has been controlled by aggressive gravel mining (Grifﬁths, 1979; Blakely and Mosley, 1987). The overall bed slope is 0.0048. Like the Waitaki River, the Waimakariri’s headwaters extend to the main divide of the Southern Alps, and most ﬂoods are associated with spring–summer rain or rainon-snow events. At the study reach, the mean ﬂow is 120 m3/s and the mean annual ﬂood is 1520 m3/s. Unlike the Waitaki, the Waimakariri’s catchment has no signiﬁcant lake storage, thus the Waimakariri’s ﬂow regime shows a greater seasonal signal (Fig. 21.3b), has more frequent freshes (Fig. 21.2), and ﬂoods are substantially larger relative to the mean ﬂow (Fig. 21.3a). Also unlike the Waitaki, the ﬂood and sediment regimes of the Waimakariri River have not been artiﬁcially regulated, although there is some extraction for irrigation at normal ﬂows and there is pressure to ‘harvest’ ﬂood ﬂows into storage reservoirs. The Waimakariri bed material is dominantly greywacke gravel. At the study reach, the average distribution of six sieved bulk samples had an overall median size of 28 mm and contained 14% sand (Environment Canterbury, unpublished data). The median size of the gravel mode was 32 mm while that of the sand mode was 0.2 mm. Maximum clast size was 152 mm. Bed surface size (sampled by Wolman counts) varies with the depositional environment. On freshly deposited lobes and sheets, the median size ranges from 12 to 30 mm but is typically about 24 mm; in channels and on scoured bars, the median size ranges between 32 and 58 mm (Carson and Grifﬁths, 1989). The overall average median surface size, sampled over bed-wide transects, is 28 mm (Grifﬁths, 1979). The match between average surface and bulk distributions indicates minimal armouring and thus an abundant supply of bed material relative to transport capacity, which is consistent with the aggradation trend recorded from historical cross-section surveys undertaken by the river management authority (Grifﬁths, 1979; Carson and Grifﬁths, 1989). This contrasts with the extensive development of surface armour on the lower Waitaki River bed. Scattered sand patches occur on the Waimakariri bed surface, either in primary depositional sites in low velocity zones of wetted braids or reworked by wind into thin drifts on elevated areas growing grasses and lupins. The bedload through the Waimakariri study reach (as determined from analysis of historical cross-sections and gravel-extraction data between 1955 and 1983) is approximately 275,000 m3/yr (Carson and Grifﬁths, 1989). Bedload transport occurs during even the small ﬂoods or ‘freshes’ (200 m3/s and greater) that reoccur on a quasi-monthly basis, although it becomes more extensive during larger ﬂoods. During freshes, it is manifest by migrating gravel lobes within the main braids, but during bankfull ﬂoods it also involves migrating gravel sheets that may be over 100 m wide (Hicks et al., 2002). Typically, the smaller bedload-transporting freshes last 1 day, while ﬂoods may span several days.

4.3.

Methods

We conducted ﬁve surveys between February 1999 and July 2003, capturing the changes due to several ﬂoods ranging up to 1393 m3/s and numerous freshes.

570

D.M. Hicks et al.

In all cases, the remotely surveyed topography data were used to generate co-located DEMs with 1 m square grids. The ﬁrst three surveys used digital photogrammetry to map the elevation of the dry areas of riverbed. As detailed in Westaway et al. (2000, 2003) and Lane et al. (2003), water depth was classiﬁed from rectiﬁed colour aerial photographs using empirical relationships between water depth and colour. The estimated depths were combined with estimates of local water surface elevation obtained photogrammetrically to provide data on bed elevation in inundated areas. The water depth calibration datasets were collected concurrently with the photography using a small dinghy equipped with an echo-sounder integrated with an RTK GPS unit. This allowed rapid and accurate data collection (at 10 Hz and cm-scale accuracy), and also provided check data on submerged bed levels. The raw photogrammetric results required post-processing in order to remove the effect of small errors in the triangulation of each photo frame during the ‘bundle-adjustment’ process (Westaway et al., 2003). Before correction, this tended to generate an interference pattern of locally varying systematic error on grids of elevation difference, which appeared as a ‘patch-work-quilt’ effect on image maps of elevation difference. The fourth and ﬁfth surveys employed an Optech airborne laser scanner (i.e., LiDAR) to survey the dry riverbed topography. With this system, a pulsing infrared laser traversed across the ﬂight path, sampling the ground elevation, while the location of the laser system in the air, with respect to the local circuit survey datum, was found by post-processed differential GPS. The pulse rate was 5 kHz in 2000 and 25 kHz in 2003. Multiple passes in 2000 yielded an average point spacing of 1.6 m while a single pass in 2003 yielded an average point spacing of 1.3 m. The wetted channel water depths were mapped from concurrent aerial photographs as described previously, with waters-edge and water-surface laser returns used to create DEMs of the water surface. ‘Last’ returns were used to maximise the chance that the laser altimetry was from the ground, not off riverbed vegetation. Some vegetation returns could not be avoided, however, and these were ﬁltered from the dataset using automated and manual approaches. First, a numerical ﬁlter removed local high points. Second, using tools within the ArcInfo software suite, vegetation-created topography was identiﬁed off shaded relief images, masked, and replaced with a local trend surface. The LiDAR data were acquired into the same projection and above the same vertical datum as the photogrammetry. While the 2000 and 2003 LiDAR surveys were reduced to datum using different geoidal models, the theoretical difference between the two models over the study reach was found to differ by less than 0.011 m, and check elevations for 100 stable points off the riverbed showed a mean elevation difference of 0.0370.02 m, which was not signiﬁcantly different from zero at the 5% signiﬁcance level. To observe morphological change at a greater frequency than afforded by the synoptic photogrammetric or LiDAR surveys, we installed two video cameras 35 m above the riverbed on electricity pylons (Hicks et al., 2002). These, solar powered and video-linked to a computer located 4 km distant, allowed us to capture images of the river bed continuously during daylight hours at 20 min intervals over the four years

Contemporary morphological change in braided gravel-bed rivers

571

of surveys. This raw oblique imagery was converted to georeferenced orthoimagery using photogrammetric methods.

4.4.

Uncertainties

The accuracy of the gridded ground elevations, and hence of the DEMs of difference, varied between surveys and also over the study area according to whether the bed was surveyed dry or wet (Lane et al., 2003, 2004). Accuracies were determined during each survey by comparison of ground-surveyed dry and wet point elevations against those interpolated from the remotely sensed DEMs. The reliability of the photogrammetry-based DEMs improved with the third survey, in February 2000, when the acquisition scale was reduced from 1:5000 to 1:4000. This scale delivered a sharper texture off the gravel riverbed surface, improving the success rate of automated matching across stereo image pairs. For this survey, the DEM is estimated as accurate to 70.13 m (root-mean-square accuracy) in dry areas and 70.22 m in wet areas. Both LiDAR-acquired DEMs are estimated as accurate to 70.10 m in dry areas and 70.25 m in wet areas. Herein, we focus on the results of the latest three surveys because of their similar and superior accuracies. Analysis by Lane et al. (2003, 2004) showed that the uncertainty on elevation change at a point can be estimated from the uncertainty of the original elevation DEMs on a root-mean-square basis. This approach was used to determine a minimum level of elevation change detection on the DEMs of difference according to whether the elevations were dry or wet at each survey and whether determined by photogrammetry or LiDAR. These levels of detection (LOD) ranged from 70.14 to 0.18 m for dry points in both surveys (dry–dry), through 70.24–0.28 m for the same point being wet in one survey and dry in the other (dry–wet), to 70.31–0.35 m with the point wet in each survey (wet–wet). For the February–May 2000 DEMof-difference, 11% of the reach was in the wet–wet class, 25% in the wet–dry class, and 64% dry–dry. Weighted on this basis, the spatially averaged LOD for this epoch was 70.20 m. Similar ﬁgures apply to the May 2000 to July 2003 epoch. Thus a representative average level of detection is estimated as 70.2 m. We do not expect that excluding areas showing elevation change below this LOD will signiﬁcantly bias the study results, since ﬁeld observations indicated that lobe and sheet deposits were typically thicker than 0.2 m while the cut depth due to migrating and incising channels were substantially greater, typically over 1 m.

4.5.

Results

Fig. 21.7 maps the local relief and elevation changes surveyed in February and May 2000 and in July 2003. On the elevation change maps, only changes greater than the 70.2 m average LOD are coloured. As described by Hicks et al. (2002), the relief shows intricate detail of braided channels and bars at several scales. Of particular note is a broad, relatively persistent-in-time braid-belt that meanders or semi-braids around higher-relief areas of riverbed. After removal of a planar trend surface, the

572

D.M. Hicks et al.

relative relief of the February 2000 surface ranges from 2.89 m below to 2.38 m above the mean level, with a standard deviation of 0.54 m and a skewness coefﬁcient of 0.235 (reﬂecting channels). The maps of change indicate that over the 2000–2003 period, change occurred over much of the riverbed but was less common in the northwest area (top-left on Fig. 21.7), where the local relief was generally higher. As interpreted from the DEMs, photographs, and video imagery, the change arose from a variety of processes including: laterally migrating channels (identiﬁable on Fig. 21.7 by matched pairs of elongated patches of erosion and deposition); bar growth and conﬂuence-pool deepening; avulsions at a range of scales (leaving new incised channel networks and inﬁlled pre-avulsion networks); gravel sheet and lobe development and migration during larger ﬂoods; and local-scale incision of channels and chutes cut on ﬂow recessions where the water surface becomes steeply draped over bar fronts. Rundle (1985) described similar features in the nearby Rakaia River. The variability of change over the study reach and between survey epochs highlights the space–time patchiness of sediment transfers, a characteristic of braided rivers noted from previous ﬁeld and laboratory studies (e.g., Stojic et al., 1998). Summary results of the changes exceeding the average 0.2 m LOD are listed in Table 21.1. For the February to May 2000 epoch, which had one moderate-sized ﬂood event peaking at 839 m3/s and four smaller freshes peaking between 250 and

Figure 21.7. Detrended DEMs of the Waimakariri River at Crossbank surveyed in February 2000, May 2000, and July 2003, and the corresponding DEMs of difference. On the latter, net erosion and deposition are colour-coded above a level of detection of 70.2 m. River ﬂows left to right.

Epoch

Erosion area (%)a

Deposition area (%)a

Area eroded in one epoch, deposited in the other (%)

Total area eroded or deposited (%)

Volume erodedb (m3 103)

Volume depositedb (m3 103)

Average depth eroded (m)

Average depth deposited (m)

Mean bed level change over total area (m)

February 2000–May 2000 May 2000–July 2003 Net February 2000–July 2003 Cumulative February 2000–July 2003

24.2

37.0

–

61.2

368

525

0.53

0.49

0.06

29.8

35.1

–

64.8

484

593

0.56

0.58

0.04

27.5

38.8

–

66.2

450

682

0.56

0.60

0.08

22.5

35.7

29.6

87.8

a

1532 for volume of envelope of change

0.64 for average thickness of envelope of change

–

Contemporary morphological change in braided gravel-bed rivers

Table 21.1. Areas, volumes, and average depths of erosion and deposition for areas experiencing elevation change exceeding 70.2 m over inter-survey epochs, Waimakariri River at Crossbank.

Total area of riverbed is approximately 2.9 106 m2. Erosion and deposition volumes calculated by integrating the actual measured elevation changes wherever these exceeded the 0.2 m level of detection.

b

573

D.M. Hicks et al.

574

406 m3/s, 61% of the riverbed area underwent signiﬁcant (i.e., 40.2 m) erosion or deposition. A greater area underwent deposition compared to the area eroded, but the average deposit was thinner than the average erosion bite so that the net change in mean bed level was small (0.06 m). Most likely, this ﬁgure underestimates the proportion of riverbed undergoing change, since some portions of the bed mapped with non-signiﬁcant change would have experienced compensating scour and ﬁll. This unmeasured component can be expected to increase with the time span between surveys, as the chance of scour and ﬁll cycles occurring increases. This helps account for the similar extents of change that were detected over the longer May 2000 to July 2003 epoch, which included eight events over 800 m3/s, a peak discharge of 1323 m3/s, plus some 20 smaller freshes. The total changes between February 2000 and July 2003 were only slightly greater than for the last epoch, indicating that at least some areas of riverbed had to have eroded and/or deposited more than once. Indeed, the envelope of change for the two epochs combined showed that at least 30% of the bed eroded40.2 m during one epoch and ﬁlled40.2 m during the next (or vice versa), so that at least 88% of the bed area had undergone signiﬁcant change even if this was not registered by the net changes. The envelope of change showed an elevation range of 3.38 m, had an average elevation change of 0.64 m, and had a volume of 1.53 106 m3 (Table 21.1). Based purely on number of ﬂoods and freshes, the real volume of gravel mobilised during the May 2000 to July 2003 epoch was almost certainly several times that mobilised from February to May 2000. These results show that: (i) events of sub-annual frequency in the lower Waimakariri induce measurable erosion or deposition (40.2 m) across at least 60–65% of the riverbed; (ii) over three years, when the peak river ﬂow did not exceed the mean annual ﬂood discharge, at least 88% of the bed experienced these levels of erosion or deposition at least once, with almost certainly other parts experiencing multiple change events but remaining undetected; and (iii) erosion and deposition and associated braiding activity were generally focussed in a belt of relatively lower local relief. The last point means that the likelihood of bed disturbance is spatially variable and is inversely related to local relief.

4.6.

Effects on riverbed vegetation

In terms of vegetation establishment, the rate of bed turnover in the active braid belt is therefore fast enough to scour (or bury) seeding plants on a sub-seasonal basis, which is consistent with the bare gravel aspect of much of the riverbed. It is only the higher, more stable areas of riverbed where woody vegetation appears able to begin to establish. The lack of established wooded islands and points on the riverbed, however, indicates that these higher areas too must be disturbed on a perhaps 3–5 yearly basis, either by ﬂoods of similar frequency or by the cumulative work of smaller events. In terms of the dimensionless time-scale notion introduced by Paola (2001), again assuming Tveg ¼ 5 years (as with the Waitaki) and taking from the above results 65% as the minimum relative rate of bed turnover per year, then T*E3.3. This assumes

Contemporary morphological change in braided gravel-bed rivers

575

that deposition as well as bank erosion would destroy vegetation, but even taking just the erosion component so that the E/B term E0.3, then T* remains41. Thus ﬂood-frequency and bed turnover rate dominate over the characteristic vegetation time scale on the Waimakariri, permitting the river to naturally maintain a braided planform largely unconstrained by vegetation. We may conclude from the Waimakariri study that the application of new technologies such as LiDAR for mapping topography of braided river beds can, by detailing the height and extent of morphological change, help explain how ﬂood frequency controls riverbed vegetation and the persistence of physical habitat. We expect that application of the same remote-sensing technologies to a regulated river such as the lower Waitaki would show a less intense rate of turnover of riverbed topography. Indeed, a LiDAR-based baseline riverbed DEM has been created for the lower Waitaki, thus it is primed for a repeat survey.

5. 5.1.

Laboratory studies of vegetation effects on braided morphology Experimental overview

Field observations of morphological change associated with vegetation expansion such as on the Waitaki River served as motivation for a current series of laboratory studies at the St Anthony Falls Laboratory (SAFL; Tal and Paola, 2007). These are designed to examine the impacts of sustained vegetation forcing on braided morphology, how vegetation might drive a change to other planforms (e.g., anabranching, meandering), and to identify the key processes underlying these changes. The experiments are conducted in a 16 m by 2 m ﬂume, with an initial condition of a bare bed of cohesionless sand on which a braided planform has established. All experiments are driven by cycling a simple two-stage hydrograph: a 1-h high ﬂow that transports sediment and reworks the channel morphology, and a low ﬂow of several days during which there is almost no sediment transport and plants are allowed to grow. Sediment is fed at the upstream end at a constant rate during the high ﬂow. Alfalfa is used for riparian vegetation. Seeds are dispersed uniformly over the bed at low ﬂow, simulating natural colonization of emergent bars and banks. Runs differ by the number of days between high ﬂow events. 5.2.

Monitoring methods

The methods used to record and measure processes and morphological change during the SAFL experiments are almost entirely image based in order to minimize disruption to the system. They include: (i) time-lapse vertical photography, using image-analysis techniques to continuously capture changes in the pattern of ﬂowing channel, bare-sand, and vegetation during high ﬂows; (ii) water depth mapping using Rhodamine dye calibrated with water-ﬁlled, sand-bedded trays of known geometry; (iii) bed topography cross-section surveys done at 25 cm spacing along the ﬂume by projecting a laser line onto the bed and photographing the laser line with a camera

D.M. Hicks et al.

576

mounted at an oblique angle to the laser (Wilson et al., 2001); and (iv) surface ﬂow velocity measurements using particle-image-velocimetry off ﬂoating particle tracers. The sediment ﬂux out of the ﬂume is measured at regular time intervals in a box trap. This wide range of data permits quantiﬁcation of change based on multiple trends, providing greater conﬁdence than in the ﬁeld where responses of the system are often based on only one or two indexes. An additional advantage is the ability to focus on processes and examine the system’s response from several angles. For example, experiments thus far show the transition from unvegetated-braided to vegetatedmeandering as being characterised by increased sediment storage. The topography surveys locate the aggradation areas while the time-lapse imagery helps to identify the causative processes (e.g., debris jams). Thus, more than just noting a change in active braid index, it is possible to characterise the change in terms of local sediment budgeting and channel hydraulics.

5.3.

Runs and results

The ongoing experiments are focused on investigating how a braided system evolves over a long period of time (up to 138 days) – with many repetitions of the seeding–growth–ﬂood cycle. The low ﬂow and growth phases of the cycles spanned 6 full days during the ﬁrst run, while a subsequent run involved a shorter, 3-day interval (so that the ﬂood frequency was doubled). The duration and magnitude of the high discharge events and the seeding density of the vegetation were set during the ﬁrst run so that lateral migration typically represented approximately 10% of the channel width during the ﬁrst high ﬂow event following the ﬁrst low ﬂow and growth period. The same ﬂood duration and magnitude and seeding density were used in the second run. As reported by Tal and Paola (2007), in both experiments the continuous addition of vegetation coupled with only partial removal by the high ﬂow of what had become established during the low ﬂow resulted in vegetation encroachment on the braidplain. In the 6-day-interval experiment, the braided channel (Fig. 21.8a) rapidly transitioned to a single-thread system as the vegetation choked off smaller and weaker channels and the ﬂow was corralled into a single dominant channel (Fig. 21.8b). For the experiment with the 3-day-interval, a similar end-point developed but at a slower rate because more vegetation was removed per ﬂood. This slower evolution suggests that natural systems that have undergone hydrologic changes may in fact still be in transition from one state to another. In both runs, the change from braiding to a single channel planform was characterized by reductions in braiding intensity (as indexed by average number of braids), lateral mobility (as indexed by the correlation between sequentially surveyed topographic cross-section proﬁles), and width to depth ratio, and by increases in maximum scour-hole depth and channel relief. Such changes were documented previously by Gran and Paola (2001) for single cycles of vegetation growth at different densities, and are consistent with those observed on the Waitaki River, the Platte River (Eschner et al., 1983), and with trends reported from a cellular numerical model of braiding and vegetation (Murray and Paola, 2003).

Contemporary morphological change in braided gravel-bed rivers

577

Figure 21.8. Evolution of (a) un-vegetated braided morphology to (b) single-thread channel with vegetated ﬂoodplain; SAFL experiments with 6-day low-ﬂow season.

The alfalfa seeds typically germinated within 24 h of sowing at the beginning of the low ﬂow period, they developed roots and stems 1–2 cm in length and approximately 1 mm in diameter after 3 days, and after 6 days they were slightly larger and stronger and more ﬁrmly rooted. Thus for Paola’s (2001) dimensionless time-scale parameter, we estimate 6 days as the time (Tveg) it takes for the alfalfa to reach maturity. For the ﬁrst run, using the measured erosion extent during the ﬁrst high ﬂow event following the ﬁrst seeding-growth phase as the initial reference value for rate of bed turnover, then the E/B term equals 0.017 (i.e., 10% of bed width erosion averaged over the 6 day growth period) and thus T* ¼ 0.1. Since this is {1, it predicts that the riparian vegetation growth should eventually prevail over braiding tendencies, as indeed was observed in the experiment. For the second run, assuming the same erosion extent per ﬂood but at double the frequency of occurrence, then the E/B term equals 0.033 and T* ¼ 0.2, again indicating vegetation dominance and predicting a single-thread outcome. This latter result is less reliable, though, because the erosion extent during the ﬁrst ﬂood was not measured and would likely have been greater than 10% owing to the shorter growth phase – thus T* would have been closer to 1.

6.

Discussion

Our discussion focuses on three issues. First, we draw together the key ﬁndings from our ﬁeld and laboratory studies in regard to the play-off between ﬂoods and vegetation on braided riverbeds and what this means to managing/limiting morphological change in regulated rivers. Second, we discuss the need to monitor and predict morphological change within the context of water management decisions as they impact on in-stream habitat and riparian landscape. Third, we emphasize the need for accurate 3D datasets on morphological change and offer an opinion, based on our experience, on the best ways to secure them.

D.M. Hicks et al.

578 6.1.

Floods vs. vegetation

From our review we found: (i) ample evidence that riparian vegetation inﬂuences channel form by multiple processes, and (ii) through the effects on plant recruitment and survivability, ﬂow regulation can encourage riparian vegetation growth and this then has the potential to catalyse morphological change. Also, for braided rivers, physical and numerical models have both shown how, while other external factors remain unchanged, increased vegetation density will diminish braiding intensity. In real rivers, the effects of vegetation can be expected to develop in a noisy fashion – on average over time – as vegetation encroaches with seasonal thrusts and seasonal or irregular ﬂoods attempt to scour it away. We have seen that the relative time scale proposed by Paola (2001), which balances the time to vegetation maturity against the time for ﬂoods to turn over the bed, ﬁts with our observations of the braided Waitaki and Waimakariri Rivers and also with the SAFL experiments of Tal and Paola (2007). The Waimakariri River scours or ﬁlls across at least 2/3 of its 1 km-wide bed on an annual basis and probably all of its bed within the 5-year time frame that river managers consider required for the woody exotic vegetation to establish, thus Paola’s time ratio, T*, is 3.3. In consequence, the bed is bare gravel for much of the time and the braiding process is not signiﬁcantly constrained by the multiple effects of vegetation (such as increasing bank strength and concentrating ﬂows). Quite likely, this state has been enhanced by the artiﬁcial conﬁnement of the river between stopbanks and willow belts (which increases the probability that incipient islands in the braidplain will be eroded). In its natural state, the Waimakariri channel did break into several anabranches (Blakely and Mosley, 1987). In contrast, the Waitaki River’s ﬂoods appear less effective at turning over the riverbed. This must relate in part to the damped ﬂood regime associated with the upstream lakes and their control for hydropower storage (the ratio of mean annual ﬂood to mean ﬂow is 3.1 in the Waitaki whereas the ratio is 13 for the Waimakariri). The T*E0.3 value for the lower Waitaki suggests that that river is suspended part way along the path between a braiding and a single-thread meandering system, held in its present state only by the regular application of herbicide and bulldozers. The SAFL laboratory experiments show clearly that, given adequate time and highﬂow events, a braided river will evolve into a single-thread channel system when it is exposed to vegetation and has a T* value E0.1. Thus, the SAFL experimental runs may well portend the future form of the Waitaki River if the Waitaki’s vegetation management program were ever to lapse. An alternative management option for controlling vegetation in regulated rivers such as the Waitaki is with controlled ﬂood releases from upstream reservoirs, if that is possible. Thus, a better understanding of the role of ﬂood frequency can help address the question: how can a ﬁxed volume of water be delivered most effectively to the channel downstream of a dam or other hydrologic structure so that the natural morphology of the system is preserved? Speciﬁcally, what ﬂood frequencies, durations, and magnitudes are most efﬁcient at preventing vegetation from becoming ﬁrmly established? To date, the methodology for designing this aspect of artiﬁcial channel maintenance ﬂows lags well behind the design of ﬂows for ‘freshening’

Contemporary morphological change in braided gravel-bed rivers

579

channel substrate by sloughing-off algae, ﬂushing ﬁne sediment, and mobilising spawning gravels (e.g., Wilcock et al., 1996; Milhous, 1998). We conclude, then, that a key control on planform morphology in potentially braided rivers, at least in temperate climates, is the relative balance between the geomorphic work-rate done by ﬂoods in turning over the river bed and the rate at which riparian vegetation strengthens the banks against erosion and adds drag to the braidplain. Paola’s (2001) dimensionless time-scale parameter appears to be a reasonable ﬁrst-order predictor of whether ﬂoods or vegetation will achieve ascendancy, driving a river towards either braided or single-thread end-points. In reality, variability and intermittency in ﬂoods will likely mean that the relative ascendancy ﬂuctuates, so that the river planform will vary about an average state.

6.2.

Need for measuring morphological change

Our examination of the lower Waitaki illustrates that natural river responses to changed controls can be complex, fuzzy, and lagged. Complex in the sense that it is difﬁcult to isolate the effects of individual factors, including mitigation measures such as vegetation control. Fuzzy in the sense that morphological change develops through a variety of processes against a background of considerable natural variability, therefore clear trends from single indices (such as braiding index off 2D imagery) may be difﬁcult to deﬁne. Lagged in the sense that the response rate of river systems is governed by ﬂood frequency, sediment continuity, and vegetation growth and succession rates. The Waitaki experience shows also that a prime driver for studying contemporary morphological change is to help predict changes in river landscape and physical habitat, both instream and over the dry riverbed, in rivers subject to proposals involving ﬂow regulation. Ultimately, this means quantitative prediction of morphological change. At present we are some way off this ideal. Therefore, there is a need to maintain the compound assault: (i) with laboratory and exploratory numerical models to isolate and quantify effects, processes, and time-scales, and to help focus expensive ﬁeld monitoring on the key parameters; (ii) by applying and evaluating numerical models, both rule- and physics-based, with ﬁeld prototypes like the lower Waitaki and Waimakariri Rivers; and (iii) by securing high resolution 3D measurements of morphological change in ﬁeld systems. The latter datasets are essential for testing numerical models, for verifying scale issues with laboratory and numerical studies, for deﬁning the natural range of riverbed variability due to business-as-usual processes, and for capturing long-term historical change.

6.3.

Where to with 3D monitoring?

While we worked with both LiDAR and photogrammetry on the Waimakariri, our experience there points to LiDAR being the superior technology for acquiring 3D topography over km-scale areas of river. The key reasons include: an active radiation

D.M. Hicks et al.

580

source (enabling night-time acquisition and no sunlight-shadow issues); no need for ground control (although check ground data are still desirable); a ﬂatter cost per unit area for LiDAR; avoiding the problems of locally systematic errors with the bundleadjustment process in digital photogrammetry; and LiDAR, through last and ﬁrst returns and return-intensity information, delivers a richer dataset for ﬁltering-out non-ground topography over vegetation. An added beneﬁt is that LiDAR offers a way to classify ground cover and map hydraulic roughness for use with 2D and 3D hydrodynamic models (Asselman et al., 2002; Hicks et al., 2006). Furthermore, recent developments in LiDAR have moved towards using green-wavelength lasers that penetrate shallow water and directly map streambed topography (Kinzel, 2002; Millar et al., 2005). While bathymetry LiDAR units have been operated for some years in the marine realm, until recently they have not delivered the same scan density, have not been able to resolve shallow depths (0–1 m range), and have been more expensive. LiDAR scans may also be coupled with a digital large-format camera so that an ortho-rectiﬁed image is acquired at the same time, helping postsurvey interpretation, ﬁltering, and classiﬁcation. Thus the arrival of an economic, one-pass scan to survey both dry and wet river beds, map water depths, and ﬁlter and classify riparian vegetation presents an exciting future for measuring morphological change.

7.

Conclusions

Our studies of contemporary morphologic change in two large braided gravel-bed rivers, one regulated and the other not, indicate that Paola’s (2001) dimensionless time-scale parameter is a reasonable ﬁrst-order predictor of whether ﬂoods or vegetation will achieve ascendancy, driving a river towards either braided or singlethread end-points. In the case of the unregulated Waimakariri River, the frequent freshes and ﬂoods are able to turn over much of its bed within the 5-year time frame required for the woody exotic vegetation to establish. Consequently, its braided bed is bare gravel for most of the time. In contrast, ﬂoods in the regulated lower Waitaki River are less effective at turning over the riverbed, the river is less braided than it was in its unregulated state, and the present braided state appears to be maintained by virtue of artiﬁcial vegetation control operations. The SAFL laboratory experiments show clearly that, given adequate time, a braided river will evolve into a single-thread channel when its bed is invaded by vegetation and ﬂoods occur too infrequently to contain the vegetation growth. Predicting the environmental impacts of river development proposals is a prime driver for predicting river morphological change, since any changes in the river morphology will underpin changes in the river landscape and physical habitat, both instream and over the dry riverbed. Since we are not yet able to reliably quantitatively predict river morphological change, there is a need for further research in the laboratory, with numerical models, and with ﬁeld measurements. Airborne LiDAR offers several advantages over alternative technologies for acquiring 3D topography data over km-scale areas of braided river that is adequate for detecting morphologic change. The arrival of an economic, one-pass scan to

Contemporary morphological change in braided gravel-bed rivers

581

survey both dry and wet river beds, map water depths, capture digital imagery, and ﬁlter and classify riparian vegetation presents an exciting future for measuring morphological change.

Acknowledgements This work was supported in part by the Foundation for Research, Science and Technology (New Zealand) under Contract C01X0308. Investigations and remote sensing of the lower Waitaki River were supported by Meridian Energy Ltd. Ude Shankar assisted with processing the lower Waitaki remote-sensing data. Bill Mecchia and staff from Environment Canterbury provided surveying assistance on the Waimakariri River, while Environment Canterbury provided river ﬂow data, the 2003 LiDAR survey data, and unpublished data on bed-material size grading. Prabu Chandramohan and Jochen Bind, NIWA, processed the 2003 Waimakariri LiDAR survey. MT was supported in part by the STC Program of the National Science Foundation under Agreement Number EAR-0120914 and NSF Grant No. EAR0207556. We thank Chris Paola for helpful discussions and the staff and students at SAFL for help with experiments. We thank Jeremy Walsh, NIWA, for reviewing the manuscript. Permission from the American Geophysical Union to reproduce Fig. 21.4 is gratefully acknowledged.

References Ashmore, P., 2001. Braiding phenomena: statics and kinematics. In: Mosley, M.P. (Ed.), Gravel-Bed Rivers V. New Zealand Hydrological Society, Wellington, pp. 95–120. Asselman, N.E.M., Middelkoop, H., Ritzen, M.R., Straatsma, M.W., 2002. Assessment of the hydraulic roughness of river ﬂood plains using laser altimetry. In: Dyer, F.J., Thoms, M.C., and Olley, J.M. (Eds), The structure, function and management implications of ﬂuvial sedimentary systems. IAHS Publication No. 276, pp. 381–388. Blakely, R.J., Mosley, M.P., 1987. Impact of the Waimakariri River Control Scheme on the river and its environment. Water & Soil Miscellaneous Publication No. 102, Wellington. Brasington, J., Rumsby, B.T., McVey, R.A., 2000. Monitoring and modelling morphological change in a braided gravel-bed river using high resolution GPS-based survey. Earth Surf. Process. Landf. 25, 973–990. Carson, M.A., Grifﬁths, G.A., 1989. Gravel transport in the braided Waimakariri River: mechanisms, measurements and predictions. J. Hydrol. 109, 201–220. Chandler, J., Ashmore, P., Paola, C., et al., 2004. Monitoring river-channel change using terrestrial oblique digital imagery and automated digital photogrammetry. Ann. Assoc. Am. Geogr. 92, 631–644. Charlton, M.E., Large, A.R.G., Fuller, I.C., 2003. Application of airborne LiDAR in river environments: the River Coquet, Northumberland, UK. Earth Surf. Process. Landf. 28, 299–306. Coulthard, T., Lewin, J., Macklin, M.G., 2007. Non stationarity of basin scale sediment delivery in response to climate change. This volume. Coulthard, T.J., 2005. Effects of vegetation on braided stream pattern and dynamics. Water Resour. Res. 41 (W04003), 9. Eschner, T.R., Hadley, R.F., Crowley, K.D., 1983. Hydrologic and morphologic changes in channels of the Platte River basin in Colorado, Wyoming, and Nebraska: a historical perspective. U.S. Geological Survey, pp. A1–A39.

582

D.M. Hicks et al.

Fuller, I.C., Large, A.R.G., Charlton, M.E., et al., 2003. Reach-scale sediment transfers: an evaluation of two morphological budgeting approaches. Earth Surf. Process. Landf. 28, 889–903. Glova, G.J., Duncan, M.J., 1985. Potential effects of reduced ﬂows on ﬁsh habitats in a large braided river, New Zealand. Trans. Am. Fish. Soc. 114, 165–181. Goff, J.R., Ashmore, P.E., 1994. Gravel transport and morphological change in braided Sunwapta River, Alberta, Canada. Earth Surf. Process. Landf. 19, 195–212. Gran, K., Paola, C., 2001. Riparian vegetation controls on braided stream dynamics. Water Resour. Res. 37, 3275–3283. Green, J.C., 2005. Modelling ﬂow resistance in vegetated streams: review and development of new theory. Hydrol. Process. 19, 1245–1259. Grifﬁths, G.A., 1979. Recent sedimentation history of the Waimakariri River, New Zealand. J. Hydrol. (N. Z.) 18, 6–28. Hall, R.J., 1984. Lower Waitaki River: management strategy. A Waitaki Catchment Commission and Regional Water Board Report. Heritage, G., Hetherington, D., 2007. Towards a protocol for laser scanning in ﬂuvial geomorphology. Earth Surf. Process. Landf. 32, 66–74. Hicks, D.M., Duncan, M.J., Shankar, U., et al., 2003a. Project Aqua: Lower Waitaki River geomorphology and sediment transport. Appendix Z of Project Aqua Assessment of Effects on the Environment, Meridian Energy, Christchurch. Hicks, D.M., Duncan, M.J., Walsh, J.M., et al., 2002. New views of the morphodynamics of large braided rivers from high-resolution topographic surveys and time-lapse video. In: Dyer, F.J., Thoms, M.C., and Olley, J.M. (Eds), The structure, function and management implications of ﬂuvial sedimentary systems. IAHS Publication 276, pp. 373–380. Hicks, D.M., Shankar, U., Duncan, M.J., et al., 2006. Use of remote-sensing technology to assess impacts of hydro-operations on a large, braided, gravel-bed river: Waitaki River, New Zealand. In: Sambrook-Smith, G.H., Best, J.L., Bristow, C.S., and Petts, G.E. (Eds), Braided rivers: process, deposits, ecology and management. IAS Special Publication 36, Blackwell Publishing, pp. 311–326. Hicks, D.M., Westaway, R.M., Lane, S.N., 2003b. A bird’s-eye assessment of gravel movement in large braided rivers. Water Atmos. 11, 21–23. Hupp, C.R., Osterkamp, W.R., 1996. Riparian vegetation and ﬂuvial geomorphic processes. Geomorphology 14, 277–295. Jarvela, J., 2004. Determination of ﬂow resistance caused by non-submerged woody vegetation. Int. J. River Basin Manage. 2, 61–70. Johnson, W.C., 1997. Equilibrium response of riparian vegetation to ﬂow regulation in the Platte River, Nebraska. Regulated Rivers Res. Manage. 13, 403–415. Johnson, W.C., 2000. Tree recruitment and survival in rivers: inﬂuence of hydrological processes. Hydrol. Process. 14, 3051–3074. Kilroy, C., Scarsbrook, M., Fenwick, G., 2004. Dimensions in diversity of a braided river. Water Atmos. 12, 10–11. Kinzel, P.J., 2002. Evaluation of EAARL, an experimental bathymetric and terrestrial LiDAR for collecting river channel topography along the Platte River, Nebraska. 2002 Denver Annual Meeting, Geological Society of America, Paper No. 102-4. Knighton, A.D., Nanson, G.C., 1993. Anastomosis and the continuum of channel pattern. Earth Surf. Process. Landf. 18, 613–625. Knighton, D., 1998. Fluvial forms and processes: a new perspective. Arnold, London. Lane, S.N., 2000. The measurement of river channel morphology using digital photogrammetry. Photogrammetric Record 16, 937–957. Lane, S.N., Reid, S.C., Westaway, R.M., Hicks, D.M., 2004. Remotely sensed topographic data for river channel research: the identiﬁcation, explanation and management of error. In: Kelly, R.E.J., Drake, N.A., and Barr, S.L. (Eds), Spatial modelling of the terrestrial environment. Wiley, Chichester, pp. 113–136. Lane, S.N., Richards, K.S., Chandler, J.H., 1995. Morphological estimation of the time-integrated bed-load transport rate. Water Resour. Res. 31, 761–772.

Contemporary morphological change in braided gravel-bed rivers

583

Lane, S.N., Westaway, R.M., Hicks, D.M., 2003. Estimation of erosion and deposition volumes in a large gravel-bed, braided river using synoptic remote sensing. Earth Surf. Process. Landf. 28, 249–271. Lie´bault, F., Pie´gay, H., 2002. Causes of 20th century narrowing in mountain and piedmont rivers of southeastern France. Earth Surf. Process. Landf. 27, 425–444. Maloney, R.F., Rebergen, A.L., Nilsson, R.J., Wells, N.J., 1997. Bird density and diversity in braided river beds in the Upper Waitaki Basin, South Island, New Zealand. Notornis 44, 219–232. Milhous, R.T., 1998. Modelling of instream ﬂow needs: the link between sediment and aquatic habitat. Regulated Rivers Res. Manage. 14, 79–94. Millar, D., Gerhard, J., Hildale, R., 2005. Using airborne LIDAR bathymetry to map shallow water river environments. Coastal Geotools 2005, Myrtle Beach, South Carolina, http://www.fugro-pelagos.com/ lidar/lib/pres/CoastalGeoTools2005_Millar_Rivers.pdf Millar, R.G., 2000. Inﬂuence of bank vegetation on alluvial channel patterns. Water Resour. Res. 36, 1109–1118. Murray, A.B., Paola, C., 2003. Modelling the effects of vegetation on channel pattern in bedload rivers. Earth Surf. Process. Landf. 28, 131–143. Nanson, G.C., Croke, J.C., 1992. A genetic classiﬁcation of ﬂoodplains. Geomorphology 4, 459–486. Nepf, H.M., 1999. Drag, turbulence, and diffusion in ﬂow through emergent vegetation. Water Resour. Res. 35, 479–489. Paola, C., 2001. Modelling stream braiding over a range of scales. In: Mosley, M.P. (Ed.), Gravel Bed Rivers V. New Zealand Hydrological Society, Wellington, pp. 11–46. Pizzuto, J., Lewicki, M., Moglen, G., et al., 2007. Predicting watershed scale ﬂuvial and ecological responses to land use and climate changes in gravel-bedded piedmont streams of the mid Atlantic region, USA. This volume. Reinfelds, I., Nanson, G., 1993. Formation of braided river ﬂoodplains, Waimakariri River, New Zealand. Sedimentology 40, 1113–1127. Rundle, A.S., 1985. Braid morphology and the formation of multiple channels. The Rakaia, New Zealand. Zeits. Geomorph. N.F. Suppl.-Bd 55, 15–37. Simon, A., Bennett, S.J., Neary, V.S., 2004. Riparian vegetation and ﬂuvial geomorphology: problems and opportunities. In: Bennet, S.H., Collinson, J.C., and Simon, A. (Eds), Riparian vegetation and ﬂuvial geomorphology. Water Science and Application 8, American Geophysical Union, Washington, DC, pp. 1–10. Simon, A., Collinson, J.C., 2002. Quantifying the mechanical and hydrologic effects of riparian vegetation on streambank stability. Earth Surf. Process. Landf. 27, 527–546. Stojic, M., Chandler, J., Ashmore, P., Luce, J., 1998. The assessment of sediment transport rates by automated digital photogrammetry. Photogramm. Eng. Remote Sens. 64, 387–395. Tal, M., Gran, K., Murray, A.B., et al., 2004. Riparian vegetation as a primary control on channel characteristics in multi-thread rivers. In: Bennet, S.H., Collinson, J.C., and Simon, A. (Eds), Riparian Vegetation and Fluvial Geomorphology. Water Science and Application 8, American Geophysical Union, Washington, DC, pp. 43–58. Tal, M., Paola, C., 2007. Dynamic single-thread channels maintained by the interaction of ﬂow and vegetation. Geology 35, 347–350, doi:10.1130/G23260A.1. Thorne, C.R., 1990. Effects of vegetation on riverbank erosion and stability. In: Thornes, J.B. (Ed.), Vegetation and Erosion. Wiley, pp. 125–144. Tsujimoto, T., Kitamura, T., 1996. Rotational degradation and growth of vegetation along a stream. In: International Conference on New/Emerging Concepts for Rivers. Rivertech 96, Chicago, Illinois, pp. 632–657. Tsujimoto, T., Kitamura, T., 1998. Interaction between river bed degradation and growth of vegetation in gravel bed river. In: Water Resources Engineering Conference. ASCE, August 1998, Vol. 2, pp. 580–585. Westaway, R.M., Lane, S.N., Hicks, D.M., 2000. The development of an automated correction procedure for digital photogrammetry for the study of wide, shallow gravel-bed rivers. Earth Surf. Process. Landf. 25, 209–226. Westaway, R.M., Lane, S.N., Hicks, D.M., 2001. Airborne remote sensing of clear water, shallow, gravel-bed rivers using digital photogrammetry and image analysis. Photogramm. Eng. Remote Sens. 67, 1271–1281.

584

D.M. Hicks et al.

Westaway, R.M., Lane, S.N., Hicks, D.M., 2003. Remote survey of large braided, gravel-bed rivers using digital photogrammetry and image analysis. Int. J. Remote Sens. 24, 795–815. Wilcock, P.R., Crowe, J.C., 2003. Surface-based transport model for mixed-size sediment. J. Hydraul. Eng. 129, 120–128. Wilcock, P.R., Kondolf, G.M., Matthews, W.V., Barta, A.F., 1996. Speciﬁcation of sediment maintenance ﬂows for a large gravel-bed river. Water Resour. Res. 32, 2911–2921. Wilson, B.N., Leaf, R.B., Hansen, B.J., 2001. Microrelief meter for ﬁeld topography measurements. Trans. Am. Soc. Agric. Eng. 44, 289–295. Winterbottom, S.J., Gilvear, D.J., 1997. Quantiﬁcation of channel bed morphology in gravel-bed rivers using airborne multispectral imagery and aerial photography. Regulated Rivers Res. Management 13, 489–499. Young, R., Smart, G., Harding, J., 2004. Impacts of hydro-dams, irrigation schemes and river control works. Chapter 37. In: Harding, J., Mosley, P., Pearson, C., and Sorrell, B. (Eds), Freshwaters of New Zealand. New Zealand Hydrological Society and New Zealand Limnological Society, Christchurch.

Discussion by G. Heritage, D. Milan, and D. Herrington Hicks and is colleagues use aerial LiDAR to collect 3D morphological data of riverbeds to create DEM’s. They quote a vertical accuracy of 70.13 to 0.22 m in their survey of the Waimakariri. We would like to highlight the advantages of terrestrial LiDAR in obtaining greater vertical accuracy and improved spatial coverage for landform survey. We compared a DEM surface produced using Terrestrial LiDAR with a set of points surveyed independently using a Total Station, and found that over 50% of the errors are within 70.02 m (Fig. 21.9; Heritage and Hetherington, 2005). We suggest that this greater accuracy will lead to an improved understanding of the local scale processes operating within gravel-bed rivers. In particular more subtle changes in bed morphology, often the precursor to more signiﬁcant change, can be identiﬁed with greater conﬁdence. Furthermore the greater vertical accuracy and improved density of measurements clearly has implications for improving sediment budget estimates when using a DEM subtraction technique and in providing accurate detailed base data for CFD modelling. Similarly the high spatial point density achieved by the technique overcomes interpolation issues and operator bias inherent in earlier DEM based studies. The utility of terrestrial laser scanning has been demonstrated through the operation of a Riegl LMS Z-210 scanning system across the Ferpe`cle and Mont Mine´

Figure 21.9. Overall accuracy of the ﬁeld laser data when compared with independent points collected through theodolite survey.

584

D.M. Hicks et al.

Westaway, R.M., Lane, S.N., Hicks, D.M., 2003. Remote survey of large braided, gravel-bed rivers using digital photogrammetry and image analysis. Int. J. Remote Sens. 24, 795–815. Wilcock, P.R., Crowe, J.C., 2003. Surface-based transport model for mixed-size sediment. J. Hydraul. Eng. 129, 120–128. Wilcock, P.R., Kondolf, G.M., Matthews, W.V., Barta, A.F., 1996. Speciﬁcation of sediment maintenance ﬂows for a large gravel-bed river. Water Resour. Res. 32, 2911–2921. Wilson, B.N., Leaf, R.B., Hansen, B.J., 2001. Microrelief meter for ﬁeld topography measurements. Trans. Am. Soc. Agric. Eng. 44, 289–295. Winterbottom, S.J., Gilvear, D.J., 1997. Quantiﬁcation of channel bed morphology in gravel-bed rivers using airborne multispectral imagery and aerial photography. Regulated Rivers Res. Management 13, 489–499. Young, R., Smart, G., Harding, J., 2004. Impacts of hydro-dams, irrigation schemes and river control works. Chapter 37. In: Harding, J., Mosley, P., Pearson, C., and Sorrell, B. (Eds), Freshwaters of New Zealand. New Zealand Hydrological Society and New Zealand Limnological Society, Christchurch.

Discussion by G. Heritage, D. Milan, and D. Herrington Hicks and is colleagues use aerial LiDAR to collect 3D morphological data of riverbeds to create DEM’s. They quote a vertical accuracy of 70.13 to 0.22 m in their survey of the Waimakariri. We would like to highlight the advantages of terrestrial LiDAR in obtaining greater vertical accuracy and improved spatial coverage for landform survey. We compared a DEM surface produced using Terrestrial LiDAR with a set of points surveyed independently using a Total Station, and found that over 50% of the errors are within 70.02 m (Fig. 21.9; Heritage and Hetherington, 2005). We suggest that this greater accuracy will lead to an improved understanding of the local scale processes operating within gravel-bed rivers. In particular more subtle changes in bed morphology, often the precursor to more signiﬁcant change, can be identiﬁed with greater conﬁdence. Furthermore the greater vertical accuracy and improved density of measurements clearly has implications for improving sediment budget estimates when using a DEM subtraction technique and in providing accurate detailed base data for CFD modelling. Similarly the high spatial point density achieved by the technique overcomes interpolation issues and operator bias inherent in earlier DEM based studies. The utility of terrestrial laser scanning has been demonstrated through the operation of a Riegl LMS Z-210 scanning system across the Ferpe`cle and Mont Mine´

Figure 21.9. Overall accuracy of the ﬁeld laser data when compared with independent points collected through theodolite survey.

Contemporary morphological change in braided gravel-bed rivers

585

proglacial environments, facilitating the study process form relationships of braided rivers, in response to the diurnal ﬂuctuations in the discharge hydrograph, and associated ﬂuctuations in sediment supply (Hetherington et al., 2005). Rapid channel changes across a 4000 m2 study reach were recorded from multiple scan points on a daily basis generating meshed scanpoint clouds consisting of between 3.5 and 4.5 million points with a mean point spacing of 0.02 m across the central active region. This allowed changes at the sub barform scale to be distinguished for the entire active ﬂoodplain and has provided a valuable insight into morphologic dynamics for these outwash systems.

References Heritage, G.L., Hetherington, D., 2005. Notes on the performance of side scanning lidar across varied terrain. International Association of Hydrological Scientists Red Book Publication IAHS Publ. 291, 269–277 Hetherington, D., Heritage, G.L., Milan, D.J., 2005. Reach scale sub-bar dynamics elucidated through oblique lidar survey. International Association of Hydrological Scientists Red Book Publication IAHS Publ. 291, 2005, 278–284.

Reply by the authors We are grateful to Heritage et al. for the discussion on terrestrial LiDAR (sometimes known as 3D laser scanning). Since writing our manuscript, we have had the opportunity to gain experience with this exciting new technology, and we agree that it provides a means to rapidly collect topography data at vertical accuracies and spatial densities that are superior to those delivered by airborne platforms. This presents the advantage that it can capture the detail of geomorphic processes and features that may lie within the error band delivered by the aerial platform; moreover, it is less expensive to acquire and deploy. Also, the centimetre-scale resolution means that the technology offers an opportunity to measure bed surface grainsize of cobble bed channels as well as their morphology. We believe, however, that this technology is another tool to add to the bag of existing tools, since it does have limitations relating to scale, dealing with wetted channels, and riparian vegetation. The ideal application area for 3D laser scanning is, in our view, a mainly dry and un-vegetated reach of river channel with dimensions at the sub-km scale. For larger reaches, it is necessary to shift the scanner – which typically has a scanning radius of several hundred metres – around a ‘circuit’, and the ﬁeld effort grows in proportion to the area of coverage. We suspect that beyond a threshold of perhaps one-several km, an airborne laser-scanning platform remains the pragmatic option. For example, the reach of the Waimakariri River that we surveyed with airborne LiDAR covered an area of 4 km2, which is 1000 times larger than the study reach described by Heritage et al. With regard to wetted channels, as far as we are aware existing 3D scanners use infra-red or near infra-red laser beams that do not penetrate water, thus they can

Contemporary morphological change in braided gravel-bed rivers

585

proglacial environments, facilitating the study process form relationships of braided rivers, in response to the diurnal ﬂuctuations in the discharge hydrograph, and associated ﬂuctuations in sediment supply (Hetherington et al., 2005). Rapid channel changes across a 4000 m2 study reach were recorded from multiple scan points on a daily basis generating meshed scanpoint clouds consisting of between 3.5 and 4.5 million points with a mean point spacing of 0.02 m across the central active region. This allowed changes at the sub barform scale to be distinguished for the entire active ﬂoodplain and has provided a valuable insight into morphologic dynamics for these outwash systems.

References Heritage, G.L., Hetherington, D., 2005. Notes on the performance of side scanning lidar across varied terrain. International Association of Hydrological Scientists Red Book Publication IAHS Publ. 291, 269–277 Hetherington, D., Heritage, G.L., Milan, D.J., 2005. Reach scale sub-bar dynamics elucidated through oblique lidar survey. International Association of Hydrological Scientists Red Book Publication IAHS Publ. 291, 2005, 278–284.

Reply by the authors We are grateful to Heritage et al. for the discussion on terrestrial LiDAR (sometimes known as 3D laser scanning). Since writing our manuscript, we have had the opportunity to gain experience with this exciting new technology, and we agree that it provides a means to rapidly collect topography data at vertical accuracies and spatial densities that are superior to those delivered by airborne platforms. This presents the advantage that it can capture the detail of geomorphic processes and features that may lie within the error band delivered by the aerial platform; moreover, it is less expensive to acquire and deploy. Also, the centimetre-scale resolution means that the technology offers an opportunity to measure bed surface grainsize of cobble bed channels as well as their morphology. We believe, however, that this technology is another tool to add to the bag of existing tools, since it does have limitations relating to scale, dealing with wetted channels, and riparian vegetation. The ideal application area for 3D laser scanning is, in our view, a mainly dry and un-vegetated reach of river channel with dimensions at the sub-km scale. For larger reaches, it is necessary to shift the scanner – which typically has a scanning radius of several hundred metres – around a ‘circuit’, and the ﬁeld effort grows in proportion to the area of coverage. We suspect that beyond a threshold of perhaps one-several km, an airborne laser-scanning platform remains the pragmatic option. For example, the reach of the Waimakariri River that we surveyed with airborne LiDAR covered an area of 4 km2, which is 1000 times larger than the study reach described by Heritage et al. With regard to wetted channels, as far as we are aware existing 3D scanners use infra-red or near infra-red laser beams that do not penetrate water, thus they can

586

D.M. Hicks et al.

only resolve the water surface topography. Hence, some alternative method is still required to survey the beds of wetted channels. While there is sophisticated software available to edit vegetation returns from the 3D point clouds, this requires a skilled operator and becomes increasingly difﬁcult as the density of riparian vegetation increases, particularly over wide ﬂat areas of riverbed or ﬂoodplain where the scanning angle is necessarily low. In those situations, ALS has the advantage that there is only one layer of vegetation to penetrate.

Gravel-Bed Rivers VI: From Process Understanding to River Restoration H. Habersack, H. Pie´gay, M. Rinaldi, Editors r 2008 Elsevier B.V. All rights reserved.

587

22 The ﬂoods of August 22–23, 2005, in Switzerland: some facts and challenges Martin Jaeggi

Abstract Just a few days before Gravel Bed River Workshop (GBR) VI, a very severe ﬂood hit large parts of Switzerland. Strong erosion occurred in several steep, apparently stable reaches of mountain streams, which produced heavy sediment loads. Where these were deposited, they caused channel avulsion or channel obstruction with subsequent inundations. In other cases, inundation was caused by the discharge exceeding the design ﬂood, which caused heavy damages in the ﬂoodplains where this case of extreme event had not been accounted for. The paper documents the oral presentation given at the conference, based on the information available at that time. 1.

Introduction

Very severe ﬂooding hit Switzerland, part of Austria and Bavaria on August 21–23, 2005, just a few days before the opening of GBR VI. The author was asked spontaneously to give a presentation at this conference about these events. In this paper, some facts about ﬂooding and morphological changes in the river valleys, as reported during the conference are presented. These descriptions were based on the documentation available to the author at that time. Therefore, this overview cannot be general. The oral presentation at that time, and therefore also the paper, focus mainly on cases that were known to the author by early September, 2005.

2.

Meteorological situation

In the days preceding the ﬂood a low-pressure cell developed over Genova inducing a strong circulation around the Alps. This brought warm and humid air from the Adriatic Sea, around the eastern side of the Alps and then brought a strong northern current to hit their northern slopes, leading to prolonged heavy rainfall. Rainfall E-mail address: [email protected] ISSN: 0928-2025

DOI: 10.1016/S0928-2025(07)11144-5

M. Jaeggi

588

Figure 22.1. Forty-eight hours cumulative rainfall, August 21–23, 2005 (courtesy of MeteoSwiss).

totals exceeded 200 mm in 48 h and 350 mm during 4 days (see Fig. 22.1). While on worldwide standards this may not be an extreme rainfall, these intensities and the resulting discharges in the rivers were extremely high, on a local standard. For many rivers, discharges exceeded a 100-year ﬂood as determined form discharge records previous to the ﬂood. Furthermore, the very high discharges and the signiﬁcant damages were not so much due to the peak rainfall intensity, but rather the long duration of heavy rainfall. This weather situation is not common in Switzerland, but according to the Swiss meteorological (MeteoSwiss) service a very similar event occurred in June, 1910.

3.

Overview

Flooding affected an unusually large area. As Fig. 22.2 shows, the area where damage was recorded extended from Lake Geneva to the Canton of Grisons, covering the whole northern slope of the Alps as well as a large part of the Swiss plateau. The following is a list of rivers in ﬂood and areas affected, although it may be incomplete.

Villeneuve (VD, Grande Eau) Lavey (VD) Sarine Thun (inundation by the lake) Bern (Aare) Interlaken Lu¨tschine Brienz (debris ﬂows, two casualties) Lucerne (lake)

The floods of August 22– 23, 2005, in Switzerland

Heavy Medium Light damages

589 Schächen r.

Lütschine r.

Bristen

Amsteg Interlaken Quelle: WSL

Reuss r. Grindelwald

Figure 22.2. Overview on ﬂood damages following the event of August 22/23, 2005, in Switzerland (courtesy of Swiss Federal Institute for Forest, Snow and Landscape Research, Birmensdorf). Also indicated are the rivers and locations described more in detail in the following text.

4.

Sarnen (OW; lake and Melchaa) Engelberg (road and railway cut for weeks) Bienne (Aare AG; lake) Emme Kleine Emme (Wolhusen, Emmenbru¨cke) Reuss (LU, AG) Thur To¨ss Muota Canton of Uri (details follow) Canton of Glarus (Linth) Weesen (SG) Klosters (GR) Susch (GR); Inn river

An example of channel avulsion

As Fig. 22.3 shows, a very spectacular channel avulsion occurred in the Bristen (Canton of Uri). Very probably, in the steep reach above the village a lot of sediment was eroded and then deposited in the ﬂatter part, where the valley is wider. The river left its original channel, which was completely ﬁlled up with sediment, and then cut a completely new channel. Unfortunately, several houses were in the way and destroyed. In Fig. 22.3 the old course of the river can be seen on the right.

M. Jaeggi

590

Figure 22.3. Channel avulsion in Bristen. Old channel ﬁlled by sediment, new channel with destroyed houses (courtesy of Canton of Uri).

5.

Emergency services and interventions

Emergency services in Switzerland are normally organised on a local level. Fire brigades, very often composed of volunteers, are not only ﬁre-ﬁghters, but also trained to intervene during ﬂoods. Generally speaking, these services worked very well during this event and helped in particular to keep the number of casualties down. Evacuations that were organised in time proved to be very important, particularly in the village of Brienz, where debris ﬂow destroyed several houses. A rather peculiar counter-example is shown in Fig. 22.4. In the village of Amsteg (Canton of Uri) the Reuss River was exceedingly overloaded by sediment from a tributary. The driver of an excavator was ordered to move into the river and to dredge the excess sediment. Once he was there he was soon in trouble and had to be saved by helicopter. The excavator, of course, was lost. This example shows that despite all the skills and the local knowledge, better instruction of the emergency people concerning the processes which are to be expected in a rare ﬂood, is necessary.

6.

The Lu¨tschine River

The Lu¨tschine River (Schwarze Lu¨tschine in the upper course) caused many problems between the well-known resorts of Grindelwald and Interlaken. In the upper part, there is succession of steeper and ﬂatter reaches. Strong erosion occurred in the steeper zones, causing bank collapse which in turn destroyed roads, railway lines and bridges. In the ﬂatter reaches strong deposition caused ﬂooding and local lateral

The floods of August 22– 23, 2005, in Switzerland

591

Figure 22.4. Amsteg. Excavator which was sent in to the Reuss River to ﬁght sediment deposition, and which had to be abandoned (courtesy of Canton of Uri).

erosion with subsequent damage. Fig. 22.5 shows the activation of a meander bend, where the railway and the road leading to Grindelwald were cut. Fig. 22.6 shows a road bridge which was mainly destroyed because deposition induced ﬂow to attack the right abutment. Since this was not deeply founded the bridge collapsed. In the lower reach the Lu¨tschine River crosses its own alluvial plain to ﬁnally reach the Lake Brienz. Here, the river is channelised and the resort of Interlaken and its surroundings protected by levees. The ﬂow far exceeded the design ﬂood, resulting ﬁrst in overtopping and then breaching of the levees. Water spilled over a wide area causing extensive damage. Fig. 22.7 shows the ﬂooded alluvial plain and in particular the motorway bypass, completely inundated. Fig. 22.8 shows the channelised Lu¨tschine River, which in the past was pushed to the right (eastern) side of the plain by river training works. In the particular reach shown by Fig. 22.8, breaching and damage could be prevented by loading the outside slope of the levee. But more upstream overtopping and then a breach occured. 7.

Reuss and Scha¨chen rivers

Fig. 22.9 shows the conﬂuence of the Reuss and its tributary, the Scha¨chen River. This carried a lot of coarse sediment up to stone and block size. This load was almost completely deposited at the conﬂuence, where the main river Reuss was not able to carry all this material away. The deposition then advanced up the lined channel, ﬁlling the channel and causing ﬂooding of the neighbouring areas.

592

M. Jaeggi

Figure 22.5. Meander migration and bank erosion following destabilisation of boulder structures in a steep reach of the Schwarze Lu¨tschine, destroying road and railway line to Grindelwald (Bernese Oberland) (courtesy of Peter Wyss, Maetzener & Wyss, Interlaken).

Figure 22.6. In the reach downstream of the one shown in Fig. 22.5, sediment deposition caused channel migration and bridge destruction (courtesy of Peter Wyss, Maetzener & Wyss, Interlaken).

The floods of August 22– 23, 2005, in Switzerland

593

Figure 22.7. Flooding of the Interlaken region by the Lu¨tschine River (courtesy of Peter Wyss, Maetzener & Wyss, Interlaken).

Figure 22.8. The Lu¨tschine River is channelised on the right edge of its ﬂoodplain. The levees were overtopped and a breach occurred. Here, temporary deposits on the outer slope helped to prevent a failure.

594

M. Jaeggi

Figure 22.9. Conﬂuence of the Reuss and Scha¨chen (from the left) river. An important volume of very coarse sediment carried by the tributary was deposited in the Reuss River. Extensive ﬂooding of the industrial zone in the background (courtesy of Canton of Uri).

In the lower part of the Scha¨chen, on its alluvial fan, the channel was lined in masonry after a similar event occurred in 1910. During that event, deposition and ﬂooding were concentrated at the top of the alluvial fan, 1.5 km upstream of the conﬂuence with the Reuss River. It was then hoped that all the bedload material would be easily moved away by the Reuss. Obviously, as the recent event showed, the transport capacity of the Reuss River was, however, substantially overestimated. In 1977 an important ﬂood occurred in the Scha¨chen with a peak ﬂow estimated to be 110 m3/s. In 2005 a peak ﬂow of 150 m3/s was recorded. More important was the fact that the discharge exceeded 100 m3/s more than 12 h. This shows how difﬁcult it still is to deﬁne a reasonable design ﬂood, even from long records, when between rare major events the discharge remains small, and when the duration of extreme ﬂoods is also important. Fig. 22.10 shows the consequences of the deposition spreading upstream. The channel was completely ﬁlled, and then water and sediment spilt into the neighbouring industrial area. The section below the railway bridge was also ﬁlled. In Fig. 22.11 a hydraulic jump is visible. It indicates the position to which the deposition front moved upstream. Deposition was as deep as the channel walls in the lower part, and approximately two-thirds of the channel depth near the hydraulic jump. Fig. 22.12 shows the blocking of a drainage canal, passing in a tunnel underneath the lined channel of the Scha¨chen River, by sediment spilling from the river channel. The ﬂoodplain upstream of the alluvial fan of the Scha¨chen River is lower than the beds of both rivers. That is why the tunnel was built. Because of the blockage another big industrial area was ﬂooded for days (see Fig. 22.9, background).

The floods of August 22– 23, 2005, in Switzerland

595

Figure 22.10. A deposition front moved from the conﬂuence upstream into the lined channel of the Scha¨chen River and ﬁlled it completely, which caused the ﬂooding of the adjacent industrial area (courtesy of Canton of Uri).

The arrangement of the tributary and a drainage canal and the vulnerability shown by the event was inherited from the engineers who designed the ﬁrst lined channel. It may therefore take decades before ﬂood protection works are really tested by an extreme event and the consequences of poor planning or design may not be apparent for 100 years or more. After the 1977 event a sediment detention basin was built. Unfortunately, it could not be placed near the apex of the alluvial fan, where it would normally be positioned to mitigate the problem. For construction reasons it had to be moved 1.5 km upstream. It is in the intermediate reach that the tributary incised in its bed and picked up 50,000 m3 of very coarse sediment. This volume was enough to start the sedimentation problem at the conﬂuence. Incision was between 1 and 2.5 m causing some supporting walls to collapse (Fig. 22.13). The bed was covered by very coarse boulders (Fig. 22.14) and one might expect full stability of the bed, but even here there was incision of 1 m. Fig. 22.15 shows the detention basin. The detention volume was far less than expected. This was partly due to the size of the bottom outlet, which proved to be too large. However, when the design was made it was taken for granted that the bottom outlet would be blocked by ﬂoating debris in case of major events. Although this outlet was 3.3 4.5 m wide, the trees were sucked by the current through the bottom outlet without blocking it. The dam was conceived as an open slit dam in 1980, when the project was set up. It was then very fashionable to favour open dams. The idea was that when the peak ﬂow was reached, the excess material could be held back, and then during receding

596

M. Jaeggi

Figure 22.11. The hydraulic jump indicates where the deposition front stopped. Upstream, all the supplied sediment was transported without problems (courtesy of Canton of Uri).

ﬂow sediment again be released. It was thought that this could prevent erosion in the downstream reach and lower the maintenance costs. During this event, however, the release of sediment after the passage of the peak ﬂow proved to be an aggravating factor. The lined channel in the lower part of the river was then already ﬁlled up by the sediment picked up in the intermediate reach. So, the sediment released by the

The floods of August 22– 23, 2005, in Switzerland

597

Figure 22.12. Just upstream of the conﬂuence, the canal draining the area between the two rivers passes in a tunnel built under the lined channel of the Scha¨chen River. Water and sediment spilled from the Scha¨chen into this drainage canal and blocked it. This in turn caused the ﬂooding visible in the back of Fig. 22.9 (courtesy of Canton of Uri).

Figure 22.13. Bank collapse following channel incision in the steeper part of the Scha¨chen River.

598

M. Jaeggi

Figure 22.14. Apparently stable bed in the steeper part of the Scha¨chen River, where large boulders structure the bed. However, incision amounted between 1 and 2 m and produced the volumes of coarse sediment causing the damages downstream.

Figure 22.15. Open sediment detention dam. It did not retain as much sediment as expected because the bottom outlet was not clogged by ﬂoating debris as expected. The release of sediment during receding ﬂood aggravated damages downstream (courtesy of Canton of Uri).

The floods of August 22– 23, 2005, in Switzerland

599

dam now also ended up in the channel already blocked, and also continued to ﬂood the industrial area beside the river channel.

8.

Positive examples

Although some protection systems failed, there are also positive examples. The restoration of the Reussdelta (see Jaeggi, 1986) was one of the earlier examples of an effort to give the river a more natural aspect. Originally, the river was led in a 300 m long canal into the Lake Lucerne. This was replaced by a natural delta designed based on a model study. The event of August, 2005 was the toughest test for the restored delta. The performance was fully satisfactory and consistent with the model tests. Fig. 22.16 shows this delta during the ﬂood. Some ﬂooding occurred on the shores, but this was only a consequence of the very high lake level. The ﬂood protection scheme on the Reuss River, between the conﬂuence with the Scha¨chen and the Reuss delta, which was set up after the 1987 ﬂood, also worked very well. In particular, a reach of a levee which is designed to be overtopped during rare events came into operation for the ﬁrst time. As Fig. 22.17 shows, some erosion occurred. But, as planned, the stepped spillway came into operation and energy dissipation was perfect. The event was also the ﬁrst test for a newly designed channel of the Dorfbach in Sachseln, after the ﬂooding of 1997. The channel was shifted away from the centre of the village through a bypass. Just upstream of the new mouth into the lake, there is a railway bridge. It was conceived in a way to allow the ﬂow to be pressurised. This is

Figure 22.16. The delta of the Reuss River, restored in 1991. Inundation of the shores was a consequence of the high lake level (courtesy of Canton of Uri).

M. Jaeggi

600

Figure 22.17. As planned, the Reuss River overtopped the levee which is protected by a stepped spillway.

exactly what happened during the event. Because the lake level was extremely high during receding ﬂow, deposition occurred in the last part of the channel (see Fig. 22.18). However, there was no overspilling of the banks or the bridge, even under these extreme conditions.

9.

Mountain stream morphology

In steep reaches of mountain streams normally there is a strong stable armour layer. Big boulders, which were washed out from older deposits like debris ﬂow fans, or fell from the rock ﬂanks, are characteristic for the bed surface. They may form step-like structures. Step-pool structures formed by the ﬂow may further increase stability (Whittaker and Jaeggi, 1982). So, during ordinary ﬂoods, despite the steep slope, this stable structure prevent bed and bank erosion. Sediment transport is supply conditioned, and does not interfere with the bed material. The transport rates are far lower than according to the theoretical transport capacity of the steep reach. During a very rare and extreme event, the boulders and the steps may be displaced by local scouring, which in turn releases ﬁner sediment from the bed and the banks. Because of the bed instability, a general widening of the river bed as well as activation of meanders follows (see also Bezzola et al., 1990). Transport rates of the eroded material may now correspond well to the theoretical transport capacity, as given for instance by Smart and Jaeggi (1983). So, there are clearly two modes of sediment transport in such mountain rivers (Koulinski, 1994).

The floods of August 22– 23, 2005, in Switzerland

601

Figure 22.18. Crossing of the new channel of the Dorfbach Sachseln by a railway line. During ﬂood, pressure ﬂow is accepted under the bridge (pressure bridge). The event caused an extremely high lake level. Pressure ﬂow was still possible despite deposition. The normal bed level is 2 m lower.

The longer the duration of a stable period of the lower ﬂoods has been, the more the morphological changes are surprising to the local people. In the case of the ﬂoods of the Lu¨tschine, it may well be that such an extreme ﬂood, able to break the armour layer in the steep reach, had not occurred since about 1850. Such rivers may be compared to dormant volcanoes (Jaeggi, 1995). The threshold discharge for which the armour layer breaks and high sediment rates are released may be estimated by the procedures of Whittaker et al. (1988), or Egashira and Ashida (1991).

10.

Flood occurrence and design ﬂood

In case of the Lu¨tschine River a gaging station exists with an 84-years record. Before the 2005 ﬂood a 100-year ﬂood of 195 m3/s was extrapolated. A 200-year ﬂood was estimated at 208 m3/s. During the event the gaging station stopped recording at 227 m3/s. The peak is now estimated to have been slightly above 250 m3/s. So it is not surprising that overtopping of the levees resulted. Since the devastating ﬂoods of 1987 in Switzerland, a change in ﬂood-defence policy was introduced (see Jaeggi and Zarn, 1990). The guidelines of the responsible service deﬁne the new policy (FOWG, 2001). Value of the land and the objects existing in the threatened area are considered when deﬁning a design ﬂood. Design

M. Jaeggi

602

ﬂood is not taken as an upper limit of consideration, but it is just the discharge up to which no damage at all is supposed to occur. If the design ﬂood is exceeded, this is a situation which has to be accepted and the ﬂood protection system should be able to cope with it. This can for instance be done, if an excess ﬂow can be evacuated over side-weirs, in order to spare the erodable levees downstream. Levee breaches and other collapses should not be accepted, even at extremely rare events. Flood management should furthermore limit damages in the areas which are not protected for extreme events exceeding the design ﬂood. This could be done by providing a ﬂood corridor, in a zone where minor damages can be accepted. Of course, this is sometimes very ambitious and it is not easy to meet all the goals, specially when funding is cut.

11.

Conclusions

There is now a lot of work being undertaken to study the different events which happened on August 20–23, 2005. The problems related to steep mountain stream morphology, and the problem of the suitable choice of the design ﬂood, as well as the planning of eventual consequences of a ﬂood exceeding it, are basically known, but have to be investigated further. These problems were dominant factors of these events. Up to now, no conclusion has been made concerning a recurrence period for these ﬂoods. At many places the peak ﬂow exceeded by far the design ﬂoods. It is not only the value of the peak ﬂow which is important, but also the long duration of high ﬂows, resulting in huge sediment loads. Sometimes up to 16 h of extremely high ﬂow was recorded. If ﬂood protection systems fail there is increasingly greater damage as ﬂoodplain development expands, for housing, transportation, industrial development or commerce. If extreme events are absent for decades, land use may have dramatically changed between events. Even long discharge measurement records are insufﬁcient to give a precise answer concerning recurrence of rare ﬂoods and to deﬁne a design ﬂood properly. Global warming may be another cause of uncertainty, since it may favour extreme ﬂoods. However, since events like the one described may occur with an unknown recurrence, global warming is then just another factor of uncertainty for a problem, which contains a lot of uncertainties, by deﬁnition.

References Bezzola, G.R., Kuster, P., and Pellandini, St., 1990. The Reuss River Flood of 1987 – Hydraulic Model Tests and Reconstruction Concepts, International Conference on River Flood Hydraulics, Wallingford, UK, 17–20 September, 1990, Paper J2, pp. 317–326. Egashira, S., Ashida, K., 1991. Flow resistance and sediment transportation in streams with step pool morphology. In: Armanini, A. and di Silvio, G. (Eds), Fluvial Hydraulics of Mountain Regions. Springer-Verlag, Berlin. Federal Ofﬁce of Water and Geology (FOWG), 2001. Guideline for ﬂood protection at rivers and streams, guidelines of the FOWG (available at: http://www.umwelt-schweiz.ch/buwal/shop/ﬁles/pdf/ phpaZfqBv.pdf).

The floods of August 22– 23, 2005, in Switzerland

603

Jaeggi, M., 1986. Non-conventional solution for river mouth design. J. Hydraul. Eng. 112 (1), 14–26. Jaeggi, M. and Zarn, B., 1990. A New Design Policy in Flood Protection Schemes as a Result of the 1987 Flood in Switzerland, Proceedings of International Conference on River Floods Hydraulics, Wallingford, UK, September 1990. Jaeggi, M.N.R., 1995. Sediment Transport in Mountain Rivers – A Review, Proceedings of the International Sabo Symposium, Tokyo, Japan, August. Koulinski, V., 1994. Etude de la formation d’un lit torrentiel, Etudes e´quipements pour l’eau et l’environnement, Centre national du machinisme agricole, du ge´nie rural, des eaux et des foreˆts, Grenoble. Smart, G.M. and Jaeggi, M., 1983. Sediment transport in steilen Gerinnen. Sediment transport on steep slopes. Mitteilung der Versuchsanstalt fu¨r Wasserbau, Hydrologie und Glaziologie der ETH Zu¨rich, Nr. 64. Whittaker J.G., Hickman, W.E., and Croad, R.N., 1988. Riverbed Stabilisation with Placed Blocks, Central Laboratories Report 3-88/3, Hydraulics Section, Central Laboratories Works Corporation, Lower Hutt, NZ. Whittaker, J.G., Jaeggi, M., 1982. Origin of step-pool systems in mountain streams. J. Hydraul. Div. ASCE 108 (HY 6), 758–773.

Discussion by Gary Williams

You showed an example of an open ‘ﬂushing’ check dam, which was ﬁlled up by the ﬂood peak and then was ﬂushed out over the recession, overloading the downstream reach. The dam does seem to have operated successfully, as surely without the dam the downstream reach would have been more overloaded, mainly during peak ﬂows. All such structures have a design capacity, and if the event is large enough to exceed it, then downstream damage etc. will occur. We have to accept the above design damage. It is not economically possible to prevent damage in all events. The design aim should be for a consistent standard down the system. Some consideration of what happens in over design events may still, though, be worthwhile – for hazard planning and storm warning and response.

Reply by the author

Sediment detention occurred during peak ﬂow as desired. The storage capacity was not fully used because the bottom outlet was larger than planned and in particular because clogging by ﬂoating debris did not occur, as it had been expected when the project was established. Very large trees had been sucked through an opening of 3.3 times 4.5 m. The contribution of sediment detention was minimal since bed erosion downstream of the dam was intense and sufﬁcient to cause the high damages presented in the paper. The release of sediment after the passage of the peak ﬂow increased the downstream problems. This sediment arrived in the downstream reach where the channel had already been heavily overloaded. In the future, a gate or a similar device will have to be installed to prevent a similar process.

The floods of August 22– 23, 2005, in Switzerland

603

Jaeggi, M., 1986. Non-conventional solution for river mouth design. J. Hydraul. Eng. 112 (1), 14–26. Jaeggi, M. and Zarn, B., 1990. A New Design Policy in Flood Protection Schemes as a Result of the 1987 Flood in Switzerland, Proceedings of International Conference on River Floods Hydraulics, Wallingford, UK, September 1990. Jaeggi, M.N.R., 1995. Sediment Transport in Mountain Rivers – A Review, Proceedings of the International Sabo Symposium, Tokyo, Japan, August. Koulinski, V., 1994. Etude de la formation d’un lit torrentiel, Etudes e´quipements pour l’eau et l’environnement, Centre national du machinisme agricole, du ge´nie rural, des eaux et des foreˆts, Grenoble. Smart, G.M. and Jaeggi, M., 1983. Sediment transport in steilen Gerinnen. Sediment transport on steep slopes. Mitteilung der Versuchsanstalt fu¨r Wasserbau, Hydrologie und Glaziologie der ETH Zu¨rich, Nr. 64. Whittaker J.G., Hickman, W.E., and Croad, R.N., 1988. Riverbed Stabilisation with Placed Blocks, Central Laboratories Report 3-88/3, Hydraulics Section, Central Laboratories Works Corporation, Lower Hutt, NZ. Whittaker, J.G., Jaeggi, M., 1982. Origin of step-pool systems in mountain streams. J. Hydraul. Div. ASCE 108 (HY 6), 758–773.

Discussion by Gary Williams

You showed an example of an open ‘ﬂushing’ check dam, which was ﬁlled up by the ﬂood peak and then was ﬂushed out over the recession, overloading the downstream reach. The dam does seem to have operated successfully, as surely without the dam the downstream reach would have been more overloaded, mainly during peak ﬂows. All such structures have a design capacity, and if the event is large enough to exceed it, then downstream damage etc. will occur. We have to accept the above design damage. It is not economically possible to prevent damage in all events. The design aim should be for a consistent standard down the system. Some consideration of what happens in over design events may still, though, be worthwhile – for hazard planning and storm warning and response.

Reply by the author

Sediment detention occurred during peak ﬂow as desired. The storage capacity was not fully used because the bottom outlet was larger than planned and in particular because clogging by ﬂoating debris did not occur, as it had been expected when the project was established. Very large trees had been sucked through an opening of 3.3 times 4.5 m. The contribution of sediment detention was minimal since bed erosion downstream of the dam was intense and sufﬁcient to cause the high damages presented in the paper. The release of sediment after the passage of the peak ﬂow increased the downstream problems. This sediment arrived in the downstream reach where the channel had already been heavily overloaded. In the future, a gate or a similar device will have to be installed to prevent a similar process.

The floods of August 22– 23, 2005, in Switzerland

603

Jaeggi, M., 1986. Non-conventional solution for river mouth design. J. Hydraul. Eng. 112 (1), 14–26. Jaeggi, M. and Zarn, B., 1990. A New Design Policy in Flood Protection Schemes as a Result of the 1987 Flood in Switzerland, Proceedings of International Conference on River Floods Hydraulics, Wallingford, UK, September 1990. Jaeggi, M.N.R., 1995. Sediment Transport in Mountain Rivers – A Review, Proceedings of the International Sabo Symposium, Tokyo, Japan, August. Koulinski, V., 1994. Etude de la formation d’un lit torrentiel, Etudes e´quipements pour l’eau et l’environnement, Centre national du machinisme agricole, du ge´nie rural, des eaux et des foreˆts, Grenoble. Smart, G.M. and Jaeggi, M., 1983. Sediment transport in steilen Gerinnen. Sediment transport on steep slopes. Mitteilung der Versuchsanstalt fu¨r Wasserbau, Hydrologie und Glaziologie der ETH Zu¨rich, Nr. 64. Whittaker J.G., Hickman, W.E., and Croad, R.N., 1988. Riverbed Stabilisation with Placed Blocks, Central Laboratories Report 3-88/3, Hydraulics Section, Central Laboratories Works Corporation, Lower Hutt, NZ. Whittaker, J.G., Jaeggi, M., 1982. Origin of step-pool systems in mountain streams. J. Hydraul. Div. ASCE 108 (HY 6), 758–773.

Discussion by Gary Williams

You showed an example of an open ‘ﬂushing’ check dam, which was ﬁlled up by the ﬂood peak and then was ﬂushed out over the recession, overloading the downstream reach. The dam does seem to have operated successfully, as surely without the dam the downstream reach would have been more overloaded, mainly during peak ﬂows. All such structures have a design capacity, and if the event is large enough to exceed it, then downstream damage etc. will occur. We have to accept the above design damage. It is not economically possible to prevent damage in all events. The design aim should be for a consistent standard down the system. Some consideration of what happens in over design events may still, though, be worthwhile – for hazard planning and storm warning and response.

Reply by the author

Sediment detention occurred during peak ﬂow as desired. The storage capacity was not fully used because the bottom outlet was larger than planned and in particular because clogging by ﬂoating debris did not occur, as it had been expected when the project was established. Very large trees had been sucked through an opening of 3.3 times 4.5 m. The contribution of sediment detention was minimal since bed erosion downstream of the dam was intense and sufﬁcient to cause the high damages presented in the paper. The release of sediment after the passage of the peak ﬂow increased the downstream problems. This sediment arrived in the downstream reach where the channel had already been heavily overloaded. In the future, a gate or a similar device will have to be installed to prevent a similar process.

604

M. Jaeggi

Not the open dam, but the bed erosion in the downstream reach is an example of the above design damage. The policy in Switzerland (FOWG, 2001) also includes damage mitigation for extreme events exceeding the usual design ﬂoods, if a favourable cost-beneﬁt analysis results. To prevent future damages in the industrial areas shown in Figs. 22.9 and 22.10, additional detention dams and a ﬂoodway are proposed.

Gravel-Bed Rivers VI: From Process Understanding to River Restoration H. Habersack, H. Pie´gay, M. Rinaldi, Editors r 2008 Elsevier B.V. All rights reserved.

607

23 Reservoir operations, physical processes, and ecosystem losses Klaus Jorde, Michael Burke, Nicholas Scheidt, Chris Welcker, Scott King and Carter Borden

Abstract Large reservoirs alter and fragment river systems across the world. The effects of construction and impoundment have been widely addressed but little has been done to understand, simulate, and quantify the operational consequences of dams on downstream river channels and ﬂoodplains. Fluviomorphologic and ecohydraulics approaches help to understand how ﬁsh habitat and substrate regimes change in the channels but very little information is available to quantify spatio-temporal changes in physical condition and habitat characteristics throughout the ﬂoodplain. These ecosystem changes are usually caused by multiple alterations such as channel dredging or realignment, diking, ﬂoodplain leveling, and drainage works, etc. An example from the Paciﬁc Northwest of the US will be used to demonstrate conceptual and numerical modeling of physical processes to evaluate and quantify effects caused by reservoir operations and to predict ecological response, optimize dam operation, or evaluate restoration potentials. 1.

Introduction

Rivers and their water have been used for irrigation, navigation, energy supply, and other purposes for thousands of years. Today, more than 41,000 large dams and reservoirs exist worldwide and many more are under construction as energy demand is soaring and water becomes increasingly valuable. As a result of these activities, 60% of the 227 largest rivers in the world are longitudinally fragmented by dams, diversions, or other human made infrastructure (Revenga et al., 2000). The purpose of reservoirs in general is to capture water exceeding momentary needs by downstream users and store it for times when it can be released for a greater economic beneﬁt, such as hydropower generation or irrigation. Many reservoirs also serve for ﬂood protection as well as other types of water resource management purposes, such as low-ﬂow augmentation, drinking water supply, etc. In most cases, they are E-mail address: [email protected] (K. Jorde) ISSN: 0928-2025

DOI: 10.1016/S0928-2025(07)11151-2

608

K. Jorde et al.

multipurpose projects. The operation of a dam, however, inevitably leads to physical and ecological changes throughout waterways and their ﬂoodplains, mostly downstream of the reservoir site, as well as in the reservoir itself, and in some cases also upstream (Ward and Stanford, 1983b; Craig, 1999). Downstream impacts develop through discontinuity in downstream gradients, e.g., sediment supply, water quality and temperature, and most of all by modiﬁcation of the ﬂow regime. The effects can vary, depending on the distance to the dam and other boundary conditions. Channel erosion and armoring and therefore a change in the longitudinal proﬁle is a consequence of interrupted sediment supply. This also causes changes in the size distribution of particles. Depending on the supply from tributaries the main river will have to erode sediments from the banks to compensate for the supply limitation. Channel erosion may lead to an interruption of the lateral connectivity with the adjacent ﬂoodplain. At the same time, dams often reduce ﬂood discharges dramatically, which reduces transport capacity of the river downstream. If there is sediment supply from the banks or tributaries, channel sedimentation with ﬁne materials may also be a consequence. In general, the combination of reduced supply of coarse materials and reduced transport capacity signiﬁcantly reduces the potential for erosion and aggradation in the channel and ﬂoodplain and therefore causes a major morphological readjustment (Petts, 1984). The changes in ﬂuvial and ﬂoodplain processes that also include subsurface ﬂows and ﬂoods consequently reshape the spatial and temporal mosaic of available physical habitats characteristic of river ﬂoodplain systems (Ward and Stanford, 1983b; Richter et al., 1996; Poff et al., 1997). These primary and secondary changes have biological and socioeconomic implications. As a result of riverine activities on North American rivers, more than 20% of the freshwater ﬁsh species have become extinct or are endangered or threatened; a rate which is ﬁve times greater than terrestrial species extinction. Seventy percent of the riparian forests have disappeared and only 2% of the rivers are considered natural (Johnson et al., 2001). Worldwide, the number of large rivers that do not reach the ocean any more is on the rise and their river deltas suffer dramatic changes. As a consequence, dwindling ﬁsh populations have resulted in economic downturns in communities supported by commercial ﬁshing. The dilemma becomes clear where water users in semi-arid or arid regions demand that every drop of water should be used in support of consumptive and nonconsumptive human needs and at the same time river ecologists support the hypothesis that the complete spectrum of the hydrological regime is necessary for driving and maintaining riverine ecosystems and ﬂoodplains. The magnitude of downstream impacts associated with a particular facility depends on many factors, including facility size, operational conﬁguration (e.g., storage, run-of-river), operational strategy (e.g., hydropower production, irrigation storage, storage, and withdrawal), reservoir conﬁguration, and regional hydrology and climate (Williams and Wolman, 1984; Graf, 1999). The downstream distance that is affected by a particular facility also varies greatly. The sustainable management of a river system requires the ability to specify and quantify the extent of these alterations, the ways in which they are interrelated, and the strategies with which they can be minimized without generally eliminating the possibility of operating each and every reservoir.

Reservoir operations, physical processes, and ecosystem losses

609

The evolution of riverine ecosystem understanding has lead to a consensus that the spatial and temporal heterogeneity of rivers can be viewed as landscape ecologists view the land, creating riverscape ecology. The extent of a riverscape is typically considered to include the channel and the surrounding riparian area to the limit of contemporary ﬂuvial processes. Riverscape ecology gives light to riverine ecology by recognizing that spatially explicit habitat types should not be viewed as independent from one another (Wiens, 2002), and provides the potential for evaluating river corridors with a holistic perspective, by rigorously integrating function, structure, and dynamics (Tockner et al., 2002). Since the early eighties, ﬂuvial ecologists have developed a whole suite of concepts to help understand the riverine ecosystem processes and functions and to a much smaller extent to develop quantitative operation strategies for mitigating the effects of dams. Concepts spanning from the River Continuum Concept (Vannote et al., 1980) to the Shifting Habitat Mosaic (Malard et al., 2006) aid river managers in understanding how rivers and ecosystems function and respond to change, but only in a qualitative way. Few available modeling tools such as RiverSmart (Egger et al., 2005) or the HGM model (Hauer and Smith, 1998) allow some assessment of the physical and ecological conditions downstream of a dam. However, most of these models are empirical and only semi-quantitative, often developed for a speciﬁc river area and they have therefore only limited capability for assessing the operational effects of dams or for optimizing mitigation efforts. Quantitative simulation modeling will be a key tool in assessing the importance of ‘natural’ riverscape character to ecologic function (Church, 2002; Richards et al., 2002; Tockner et al., 2002). The conceptual model described in this paper is based on a pilot study to address operational ecosystem losses due to a large dam and will provide a framework to quantify and isolate the effect of anthropogenic alteration on riverscape characteristics. This paper presents a methodology to quantitatively determine the effects of dam operations on downstream habitat conditions. The methodology is ﬂexible to accommodate the varied conditions found throughout the world’s river systems. In a case study, the conceptual model is coupled with numerical modeling, the methodology will (1) isolate operational impacts from other basin changes, (2) assess the manifestation of operations-based inﬂuences on downstream physical processes, (3) link physical process evaluation with biological processes and ecological function, and (4) be applicable in a predictive capacity. These qualities will allow river managers to isolate the physical and biological effects associated with the dam and develop river-management strategies to mitigate those effects. While the focus of this study is on physical and biological losses, certain socioeconomic and possible health or lifestyle implications exist as part of dam operations. The large costs of planning, construction, and maintenance of the dam puts a strain on funding organizations and companies, while local communities may need to make adjustments to their cultural or recreational uses of the river system. For optimal watershed management, all of these losses would need to be integrated into a much larger model that spans across all areas of the environment, society, and economy.

K. Jorde et al.

610 2. 2.1.

Rivers and dams River system concepts

Over the last decades a series of river system concepts have evolved that help us understanding how rivers and their adjacent ﬂoodplains function in terms of the interaction between the physical and biological features of the system. While the earlier concepts regarded rivers as linear systems from headwaters to the mouth, the later ones shifted more towards dynamic interactions on multiple spatial and temporal scales. Here, we try to give a brief overview of the most important concepts. The River Continuum Concept, proposed by Vannote et al. (1980) describes the physical changes occurring within a river channel from its headwater to its mouth where it may discharge into a lake or ocean, regarding changes in stream velocity and temperature, sediment composition and morphology, and primary nutrients shifting from external to internal sources. As an ecological response, the natural species composition is also shifting. This continuum of running water and sediment that shapes the ﬂuvial environment is interrupted as described by Ward and Stanford (1983b) in their Serial Discontinuity Concept. It describes the interruptive effects of dams and reservoirs on river systems, and in particular, ﬂows, temperature patterns and variations, substrate composition and distribution, water depth, suspended sediment content, and organic production. Later the concept of discontinuity was extended to also incorporate ﬂoodplains (Ward and Stanford, 1995). Several concepts focus on temporal variability. Originally developed for coral reefs and tropical rainforests, the Intermediate Disturbance Theory (Connell, 1978) emphasizes the role which disturbances of different types and magnitudes have for the development of biocenoses, it was later also applied to riverine systems (Ward and Stanford, 1983a). The Flood Pulse Concept (Junk et al., 1989) underlines the idea that ﬂood events drive the productivity and interactions of major organisms living in the river-ﬂoodplain system, and that most species have adapted to and depend on predictable ﬂood events that occur at regular intervals and durations for a particular area. The Patch Dynamics Concept (Townsend, 1989) assumes that, within a given area, temporal and spatial variability are strong determinants of species richness and diversity, and where the level of temporal variability provides varying advantages to differing species. As temporal variability continues to increase, only highly adaptable or mobile species may survive, reducing or eliminating the level of competition in the ecosystem. Frisell et al. (1986) established the hierarchical framework for stream habitats and pointed out that processes removed from the system or disturbed on a high hierarchical level (such as the catchment) could not be successfully compensated or ‘‘restored’’ on a lower level, such as a reach. Ward (1989) described the four dimensions over which natural river systems can be observed: longitudinal, lateral, vertical, and temporal aspects. The longitudinal dimension can be disrupted by physical barriers (such as dams); the lateral dimension considers the riparian zone and ﬂoodplain on each side of the channel; the vertical aspect consists of the interaction between the river bed, hyporheic zone, and subsurface aquifers; and the temporal aspect of this framework relates to the timing of the ﬂow regime, particularly drought or ﬂood periods and their frequencies.

Reservoir operations, physical processes, and ecosystem losses

611

Boon (1992) added a ﬁfth dimension, the conceptual one, related to the concept to be applied for river conservation. Poff et al. (1997) suggested that alterations to the Natural Flow Regime (ﬂow magnitude, frequency, duration, timing, and rate of change) can, directly or indirectly, inﬂuence secondary physical processes and thus have deleterious impacts on ecological integrity. Richter et al. (1996) proposed the index of hydrologic alteration which is a good tool to quantify or compare hydrologic alterations. However, there is little knowledge as to what quantitative components of the hydrologic regime are really important for ecosystem function, which ones are less important and which components of the natural regime can be changed and how. These above-mentioned concepts have more or less all merged together. None of them were incorrect, but none managed to describe the complete structural and functional attributes of riverine ecosystems at a range of spatial and temporal scales when considered in isolation. More recently, riverine ecosystems have been viewed more holistically, similarly to landscapes, taking patch types and their spatiotemporal context into account and thus leading to the concept of riverscapes (Wiens, 2002). Today’s understanding is that the life cycles of many riverine species rely on a shifting ﬂoodplain landscape or habitat mosaic where a dynamic riverscape sustains biodiversity by providing a variety of refugia (Robinson et al., 2002). In general, these concepts are not meant for quantitative assessments of ecosystem functions. Parallel to the conceptual models, very speciﬁc quantitative methods linking certain physical processes or parameters with ecological response have been developed for selected species or taxa, e.g., for ﬁsh, benthic organisms, and aquatic or ﬂoodplain vegetation (Akeret, 1982; Bovee, 1986; Statzner et al., 1988; Schmedtje, 1995; Jorde and Bratrich, 1998; Lamouroux et al., 1998). As our understanding of river systems has matured, we have changed how we view the river. The river can no longer be treated as a simple one-dimensional line in the landscape. Instead, it demands a much deeper understanding of its far-reaching multidimensional characteristics. Our view of the river continues to evolve where the most recent concepts investigate patch dynamics and the temporal disposition of complex riverine ecosystems. The dynamics of these patches are mostly driven by the ﬂow of water and sediments. Dams alter the river continuum, and interrupt physical and biological processes. The following chapter will describe how these impacts can be organized into orders, with higher order processes having inﬂuence over following orders.

2.2.

Implications of hydrologic interference

There is a vast amount of evidence about the impacts of large dams and reservoir operations that change the hydrological regimes (see, e.g., Petts, 1984; Ligon et al., 1995; McCully, 1996; Craig, 1999), such as changes in the river bed and ﬂoodplain structure and morphology, including deltas and estuaries, due to altered transport capacity and reduced sediment supply. Water quality aspects downstream of dams include, e.g., changes in temperature, nutrient loading, turbidity, and dissolved gases. The change in the hydrological regime usually affects all components such as

K. Jorde et al.

612

magnitude, timing, duration, return periods and those changes alter ﬂow velocities, water depths, ﬂoodplain inundation patterns, and ﬂood pulses (Junk et al., 1989; Heiler et al., 1995), but also surface/groundwater interaction (Merritt and Cooper, 2000) or the supply of large woody debris (Naiman et al., 2000a,b). Only some of those alterations can be prevented, e.g., by bottom outlets that can handle sediment ﬂushing or by multilevel intake structures (Bratrich and Truffer, 2001). This affects all habitats and therefore species composition on all trophic levels changes as well, usually resulting in a dramatic loss in diversity. In order to structure the impacts as needed for the intended modeling processes we suggest a systematic order of impacts, based on a framework originally proposed by Petts (1984) and illustrated in Fig. 23.1. The ﬁrst order of impact represents the immediate physical responses due to the presence of the dam and reservoir. Depending on the size of the reservoir and operational mode the ﬂow regime is altered, and the reservoir acts as a trap for sediments and nutrients and the water quality may change due to the residence time and processes in the reservoir. The second order is also composed by physical processes downstream of the dam, but produced primarily by the previous order of impacts. For example, the altered hydrology leads to hydraulic changes including altered water depths, velocities, shear stress, and sediment size and composition. This also leads to the abovementioned changes in channel and ﬂoodplain morphology. The third order relates to biological impacts or ecological response, e.g., the composition of ﬁsh communities, due to the physical changes indicated primarily by the second order. Finally, the fourth order is the biological feedback that may alter the second and third orders. Examples of fourth order impacts are changes in ﬂoodplain roughness due to vegetation growth that then will change the hydraulics of the affected river

1st Order Upstream Conditions

2nd Order Physical Model

3rd Order Biological Model

Hydrology

Floodplain Morphology

Water Quality

Channel Morphology

Floodplain & Aquatic Vegetation

4th Order Feedback

Figure 23.1. Order of impact schematization.

Sediment Supply

Channel & Floodplain Hydraulics

Sediment Transport Capacity

Invertebrates, Fish, Birds, & Mammals

Biological Feedback

Reservoir operations, physical processes, and ecosystem losses

613

reach, or the loss of large woody debris, which changes channel structure if riparian forest recruitment is disabled. Each of the four orders require separate modeling processes. The ﬁrst order determines the model’s physical upstream boundary conditions followed by the hydraulic and transport models in the second order. Biological simulations are performed in the third order. Finally a fourth order is a feedback process that supplies changed initial conditions to the second and third order for future simulations.

2.3.

Requirements for an integrated modeling approach

The nature of unimpaired riverine ecosystems as we understand them today is increasingly complex and based on the idea of a shifting habitat mosaic (Malard et al., 2006) in four dimensions where patches are in different successional stages depending on when the succession began, e.g., when a bar was freshly eroded or when a channel was cut off. The patches differ in size and quality, their boundaries are ﬂexible because of hydrologic and ﬂuviomorphologic processes and the qualities of the boundary ecotones depend on the characteristics of neighboring patches. The connectivity between patches is driven by hydrologic cycles (Amoros and Bornette, 2002) and by the spatial arrangement of patches within larger ones or within the ﬂoodplain. For some organisms connectivity will be good for others bad (Wiens, 2002). Scale of observation and modeling plays an important role because patterns and species response are scale-dependent. If an incorrect scale is chosen, patterns might disappear or appear to be relevant when they are not. This spatial pattern which is the basis for ecosystem functions is driven by the physical processes. In most cases, hierarchical physical processes and ecological functions must be modeled at different scales. The scale at which they are modeled must represent the scale at which processes under investigation change at relevant rates. Usually, a larger scale model will provide boundary conditions for a smaller scale (and/or more complex model). For example, 1-D hydrodynamic models are used to route ﬂoods down the river and results of those models are then used to address smaller scale processes like local shear stress or vegetation recruitment potential using 2-D hydrodynamic models. The same applies for the data required to feed the models, such as digital elevation models. While the ﬂooding of the Lower Kootenai Floodplain in the case study partially presented here is investigated with a model that has a 30 30 m grid size, we use a 10 10 m grid model to analyze vegetation habitats later on. Given the complexity of the structure and physical processes that are the foundation of riverine ecosystem it becomes quite clear that this cannot all be modeled. First, many of the processes are non-linear and not fully understood regarding their underlying mechanisms, therefore suitable models do not exist or need too much input data or have to be calibrated extensively. Second, the input data for such models are not readily available. They are inhomogeneous and anisotropic throughout the ﬂoodplain, usually remainders of many previous ﬂood events of different magnitude and duration. Remote sensing such as LIDAR (LIght Detection And

K. Jorde et al.

614

Ranging), hyperspectral imagery or ground penetrating radar may solve some of those problems but it will still remain nearly impossible to predict where exactly, e.g., a new channel will be formed through avulsion in a ﬂoodplain, because local effects that cannot be represented in large-scale models play an important role. We have two options, either to measure directly, or to model simpliﬁed processes. The question is therefore, what can we actually model and which surrogates can we use to make predictions on processes that have been identiﬁed as relevant but cannot be modeled. It is important to remember that the relative abundance of individual riverscape features remains constant (Ward et al., 2002) if situations with similar time spans since the last disturbance that alter those features are compared. Ward et al. (2002) identiﬁed, e.g., a multitude of different surface water bodies but it would be very difﬁcult to model the spatial arrangement of such water bodies within a ﬂoodplain. The processes that lead to such arrangements, however, are erosion and aggradation of solid particles, sediment and deadwood, within the ﬂoodplain and subsequent surface water and ground water ﬂuctuations. Erosion and deposition are probably driven by transport capacity over erodible areas, expressed in shear stress or stream power, and sediment supply either from the river or eroded from the ﬂoodplain itself. It appears therefore that rather simple modeling approaches, e.g., a hydrodynamic model based on a ﬂow regime and a digital terrain model, would tell us if the processes (or disturbance level) leading to some kind of desired geomorphological setting, which then supports formation of the riverscape, are present under certain conditions and to what extent.

3.

Conceptual model

The conceptual model we propose here provides a framework within which to structure, analyze and quantify the ecosystem impacts of altered hydrologic regimes below dams. We developed this conceptual model for a pilot study on the Kootenai River in northern Idaho (Fig. 23.2). It has three major purposes, ﬁrst to establish the systematic template of a speciﬁc catchment or ﬂoodplain with all contributing impacts, second to identify reference scenarios and compare those with the contemporary or other historical impacted situations, and third to allow for the evaluation of future restoration strategies. The conceptual model helps to structure the information about a given basin in order to quantify the alterations of the system, which are due to operation of the dam rather than other human impacts to the river-ﬂoodplain system. The general framework can be applied in simple situations where the dam is the only human impact as well as in situations where there are complex, cumulative impacts to the river system. Basin changes such as channel modiﬁcation and diking, ﬂoodplain simpliﬁcation and drainage, watershed land-use changes, and climate change can confound the evaluation of the downstream effects of dam and reservoir facilities (Williams and Wolman, 1984). Speciﬁc numeric models like those described at the end of this section allow for rigorous numerical simulation of selected physical alterations of river-ﬂoodplain systems and the consequent biological impacts.

Reservoir operations, physical processes, and ecosystem losses

615

Kootenai River

Study Area Canada Kootenay Lake USA

Corra Linn Dam

Ft. Steele

Queens Bay

Wardner Kuskonook Canada

Creston Ferry Porthill Copeland

USA Leonia Libby Reservoir

Bonners Ferry

Libby

Washington

Libby Dam Idaho

Meandering reach

Braided reach

Montana

100 km Canyon reach

gage location

Figure 23.2. Location map showing the Kootenai River Basin, with gaging station locations and geomorphic reaches indicated.

3.1.

Model description

River ﬂoodplain ecosystems are complex environments made up of physical and biological components which vary in space and time. The exact structure of the ecosystem in space and time is created and maintained by the hydrologic regime ﬂowing over and through the physical template. The controlling variables of a regulated system are the physical template, the hydrologic regime, and the dam operations, as shown on the horizontal axis in the upper part of Fig. 23.3. The physical template includes the river-bed bathymetry and sediment composition, the ﬂoodplain topography, surface properties such as sediment characteristics or vegetation type, and the hydraulic/ﬂuviomorphologic connection between the river and ﬂoodplain. The hydrologic regime is composed of the distribution of ﬂows in the river and the connected ﬂoodplain groundwater aquifer. Dam operations describe the alteration of the ﬂow between inﬂow and outﬂow of the dam, including normal operation as well as controlled and uncontrolled spill. Dam operations must

time

Governing Variables Possible feedback loops

Pristine 1880s: Timber harvest

1880s: Limited navigation dredging

1900-1920s: Early floodplain conversion

1920s-1930s: Diking & drainage, Idaho &British Columbia

1930s: Corra Linn completed (1932), Grohman Narrows ex.

Historic Floodplain & Channel

Pre-1938: Historic Scenario 1938: Int. Joint Commission order of approval 1940s: Major levee construction, Bonners Ferry to Int. border

1939-1967: Pre-Dam Scenario

1939: ½ Floodplain behind dikes

1917-1920s: Lake Ck, Bull and Elk Dams

No Dam

1949: Moyie Dam

1950s: Levees at Libby, Bonners Ferry

Pre-Dam Floodplain & Channel

Pre-Dam Hydrological Regime

1974-Present: Discontinuity of dissolved and particulate matter, H20 quality

Post-Dam Floodplain & Channel

Pristine Physical Processes

1910: Railroad

Historic Hydrological Regime

1967-1974: Libby Dam construction

1974-Present: PostDam Scenario

No Dam

No Dam

1970s: Implement forest practices

1974: Libby Dam online

1974-91: Libby Dam operation regime 1

Physical Process Modeling & Analysis

Pristine

Pre-1862: Pristine

Dam Operation

1991-Present: Libby Dam operation regime 2

Post-Dam Hydrological Regime

Present Libby Dam operations

Restoration actions

Future

Future Floodplain & Future Hydrological Regime Channel

Differential Evaluation of Physical Processes between Reference Scenarios

Pre-Dam Physical Processes

Historic Ecological Functions

Pre-Dam Ecological Functions

Post Dam Physical Processes

Post-Dam Ecological Functions

Future Physical Processes

Future Ecological Functions

Differential Evaluation of Ecological Function between Reference Scenarios

Figure 23.3. Schematic representation of conceptual integrated modeling framework for assessment of operations-based ecosystem losses applied to the Lower Kootenai River, including chronology of basin disturbance and delineation of reference scenarios.

K. Jorde et al.

Quantification of ecosystem losses

Future Libby Dam Operations

Historic Physical Processes

Pristine Ecological Functions

Biological Modeling & Analysis

Hydrological Regime

616

Physical Template

Reservoir operations, physical processes, and ecosystem losses

617

also account for reduced sediment supply downstream of the dam. If water quality is considered, the inﬂuence of the reservoir must be taken into account as well. All three of these controlling variables are directly or indirectly altered by humans. The physical template can be modiﬁed through direct human intervention by diking, levees, etc. while the hydrologic regime can be altered directly or indirectly by upstream diversions, climate change, or large-scale land-use changes in the watershed. The model organizes the progressive human alteration of the river-ﬂoodplain system through time along the vertical axis of Fig. 23.3 from pristine conditions at the top, to contemporary conditions at the bottom. Each successive alteration of the system is added to the framework at the time it occurs showing how it alters the channel and ﬂoodplain conﬁguration, hydrologic regime, and/or dam operations. Fig. 23.3 shows the conceptual model developed for the Kootenai River between Libby Dam and Kootenay Lake. In the initial situation (pre-1862, top of Fig. 23.3) the physical template and the hydrological regime are pristine and there is no dam. Then, moving downwards along the time axis in Fig. 23.3, the physical template is changed by diking, dredging, surface leveling for agriculture, reduced Kootenay Lake levels because of Cora Linn operation, etc. Additionally some smaller dams on tributaries were built. In 1972, Libby Dam is added thus changing the hydrological regime. The horizontal arrows in Fig. 23.3 show the combination of different historic situations as compared to today. The modeling processes will follow these arrows, thus producing comparable sets of results describing the physical processes for different historic or present situations. Additionally, virtual situations can be constructed where a combination of today’s ﬂoodplain with the historic hydrological regime is modeled. In a further step, the results of the physical process modeling can be coupled with biological interfaces that model habitat quality for certain organisms or groups of organisms, if such interfaces are available. In a ﬁnal step, the results for different scenarios are quantitatively compared, the differences are a measure for the ecosystem functional losses.

3.2.

Scenarios

A given combination of the three controlling variables (physical template, hydrologic regime, and dam operations) determines the habitats of the channel and ﬂoodplain. In turn this particular habitat determines the ecological function or condition of the river-ﬂoodplain system. A set of controlling variables along with the resultant physical and ecological conditions are together considered a scenario. Scenarios can be chosen to represent historical conditions – usually this is the reference scenario, which determine the impact of individual alterations to the system, or evaluate future dam operations for remediation. Historical scenarios can be compared to look at the changes in the system through time. To isolate the impacts of different alterations, it may be necessary to create combinations of the three controlling variables which did not occur, e.g., looking at the impact of hydrologic alterations if levees had not been constructed. Future scenarios can be evaluated as ‘‘what-if’’ situations to determine the effectiveness of proposed remediation strategies.

K. Jorde et al.

618

Scenarios could also include situations where restoration efforts are focused on single species, such as white sturgeons (Acipenser transmontanus) in the Kootenai River, to ﬁnd out what these single target mitigation efforts do to other ecological functions.

3.3.

Differential evaluation of scenarios

For each scenario, the physical function of the river-ﬂoodplain system is deﬁned by a suite of physical parameters which are numerically modeled for the given controlling variables. These parameters can then be used in aquatic or riparian habitat models to predict the impact to speciﬁc organisms or groups of organisms. These analyses will produce a suite of biological parameters that deﬁne the ecological function. By comparing the physical and ecological parameters for respective scenarios, the impact of an operational strategy can be evaluated. For instance, by comparing the results between a pristine reference scenario and the contemporary scenario, the magnitude of total perturbation can be determined. Both the physical processes and ecological function are expected to vary along multiple axes. By comparing different scenarios we can look at historical snapshots or predict the system response to future management actions. Comparison between contemporary and future (such as integrating a proposed operational modiﬁcation or restoration action) scenarios reveals the beneﬁt of a proposed action. By comparing the results between contemporary pre-dam (i.e., integrating all basin changes except dam construction and operation) and post-dam scenarios, the magnitude of perturbation due to facility operations can be isolated. In this last comparison, patterns of physical processes and ecological functions resulting from different scenarios are compared. The results from those ‘‘differential evaluations’’ allow quantifying operations-based ecosystem losses.

3.4.

Supporting numerical and statistical models

The conceptual model provides a way of organizing the relevant information and structuring the analysis to focus on the impacts which are most important. The use of numerical models within this framework allows for an objective analysis of the impacts and remediation strategies. The choice of models will be dictated by the necessary spatial and temporal resolution of the processes and functions in question. An analysis of the spatial distribution of cottonwood recruitment on gravel bars would require two-dimensional hydrodynamic ﬂood models. On the other hand, a water quality model may simply require a one-dimensional hydrodynamic model to predict average ﬂow velocities and dispersion rates. The choice of models depends on the questions that are to be answered, the available input data, and the scale of the modeling project. In all cases, integrated modeling efforts are required that link different types of models together. The modeling process should follow the same order of impacts described before, with upstream conditions leading to physical impacts which lead to biological

Reservoir operations, physical processes, and ecosystem losses

619

impacts. In some cases there may be a ﬁnal layer of models predicting biological feedback to previous levels. The ﬁrst step in the numerical modeling process is to establish ﬂow regimes upstream and downstream of the dam and to evaluate the possibly critical changes in the ﬂow regime (Richter et al., 1996). Sediment and water quality inputs from upstream may not be available, and will therefore need to be modeled. The accuracy of these suspended sediment and water quality models can be very good with sufﬁcient data available for conditions in the reservoir. In the next step, the physical impacts of these changes have to be modeled. While the movement of water can be modeled accurately with enough data on channel conditions, ﬂoodplain topography and for calibration, the transport of sediment is far more difﬁcult to predict accurately (Church 1995; Montgomery and Bufﬁngton, 1998; Bufﬁngton and Montgomery, 1999; Rubin and Topping, 2001). Determining the movement of the entire channel is even more problematic (Huggenberger et al., 1998; Richter and Richter, 2000). If hyporheic conditions are to be modeled, coupled surface/groundwater models may be required (Vekerdy and Meijerink, 1998; Pringle, 2003). Once the physical processes are established, biological impact models are the next step. The ability of such models to predict the suitability of a habitat for speciﬁc organisms is dependent on the importance of physical habitat outweighing other limiting factors. This sort of modeling is only predicting the possibility of habitat use by organisms, and does not predict that the organisms will be present in the system if they are currently absent. There is an increasingly large number of different types of such aquatic habitat models available today, e.g., for ﬁsh (Crowder and Diplas, 2000; Leclerc et al., 2003; Schneider and Jorde, 2003) and riparian vegetation (Friedman and Auble, 1999; Merritt and Cooper, 2000; Polzin and Rood, 2000; Poole et al., 2002; Hill et al., 2003). A recent overview can be found in COST-626 (2005). One step further, physical habitat quality can be linked to population dynamics or growth models (Gouraud et al., 2004). All models will need to use a number of empirical shortcuts as deterministic understanding is limited and it is impossible to measure all of the factors that determine reproduction, growth, and survival. While the feedback of organisms has been discussed in the literature there are only a few quantitative models which predict changes to the channel or as a function of abundance or density of organisms (Ligon et al., 1995; Naiman et al., 2000b).

4. 4.1.

Case study: lower Kootenai river Basin description

The Kootenai Basin (Fig. 23.2) is an international watershed (spelled Kootenay in Canada, Kootenai in the U.S.) originating in the northern Rocky Mountains of eastern British Columbia, Canada. From its headwaters, the Kootenai River ﬂows south into Libby Reservoir whose 145-km length straddles the Canada–USA border (Hoffman et al., 2002). From Libby Dam near Libby, Montana, the river ﬂows

620

K. Jorde et al.

through Montana, Idaho and back into British Columbia before emptying into Kootenay Lake. Located 357.1 river km above the mouth, Libby Dam and Reservoir is a 130-m high ﬂood control and hydropower production facility that impounds ﬂows originating in the upper 23,300 km2 of the 50,000 km2 basin. Kootenay Lake is a naturally formed lake whose levels have been regulated following deepening of the outlet and construction of Corra Linn Dam near Nelson, British Columbia in the 1930s (USACE, 1984). Snyder and Minshall (1996) ﬁrst classiﬁed three geomorphic reaches between Libby Dam and Kootenai Lake: a canyon reach, a braided reach, and a meandering reach. The canyon reach runs between Libby Dam (river km 357.1) and river km 257, a point approximately 2 river km below the conﬂuence with the Moyie River, and is characterized by alternate conﬁned and semi-conﬁned subreaches, with pool-rifﬂe and plane bed channel morphologies (Montgomery and Bufﬁngton, 1997). Bed-surface materials in the canyon reach range from gravel to small boulders, with isolated occurrences of large colluvial boulders in conﬁned reaches. Floodplain areas adjacent to the river channel occur intermittently, but are generally limited. Sinuosity through the canyon reach is low (1.02). The braided reach is located between river km 257 and the town of Bonners Ferry, Idaho (river km 246), and is considered a transitional section between canyon and meandering reaches. The ﬂoodplain and active channel widen through this reach, and the channel is characterized as a complex of a main channel combined with a multitude of secondary channels that are engaged as water levels rise. Bed-surface materials in the braided reach are gravels and cobbles. Sinuosity increases through the braided reach (1.2), while channel gradient is reduced (0.19 m/km (Snyder and Minshall, 1996)). Beyond a bedrock constriction at Bonners Ferry (river km 246), the Kootenai enters into the historically glaciated, north–south trending Purcell Trench, through which it ﬂows to the conﬂuence with Kootenay Lake (river km 124). In this reach, the channel broadly meanders across an extensive, 5 km wide ﬂoodplain. The ﬂoodplain is constructed primarily of lacustrine deposits accreted during periods when the west arm of Kootenay Lake was dammed by glacial ice, causing the Lake to ﬂood south to Lake Pend Oreille (Kootenai Tribe of Idaho and Montana Fish Wildlife and Parks, 2004). The bed and banks are constructed of ﬁne-grained silt and sand materials. Prior to European settlement, the Kootenai constructed natural levees on the channel margins as sediment was deposited in response to reduced shear stress and increased boundary roughness when ﬂood waters breached the banks of the river. These natural levees were improved following settlement to reclaim the Kootenai bottomlands for agricultural use (Kootenai Tribe of Idaho and Montana Fish Wildlife and Parks, 2004). At the Leonia gaging station the Kootenai has a mean annual ﬂow of 390 m3/s. The wetted width is between 100 and 200 m, with narrower reaches in the canyon section and wider ones in the braided reach, and water depths are between 2 and 25 m. The natural ﬂow regime for the Kootenai River has a sustained peak in late spring, followed by a gradual recession to base ﬂow by September, and low winter ﬂows, a pattern typical of snowmelt-dominated basins in western North America.

Reservoir operations, physical processes, and ecosystem losses

621

The meandering reach is heavily inﬂuenced by hydraulic backwater from Kootenay Lake. This inﬂuence has been present both under historic conditions and following implementation of Corra Linn Dam in the 1930s. The head of the backwater is migratory, but may be located as far upstream as river km 254.3 under certain conditions. This location is in the braided reach, several river kilometers above Bonners Ferry (Berenbrock, 2005). Sinuosity increases signiﬁcantly (1.7) and channel gradient (0.006 m/km (Snyder and Minshall, 1996)) decreases signiﬁcantly through this reach. 4.1.1.

Basin-management history

The Lower Kootenai Basin has been intensively managed starting in the late 1800s. Perturbations considered most pertinent to the Kootenai River/Libby Dam assessment include: ﬂoodplain diking, drainage, and conversion to agriculture over approximately 21,000 ha between Bonners Ferry and Kootenay Lake (1900–1940s), channel dredging coincident with ﬂoodplain diking (1900–1940s), completion of Corra Linn Dam (1932), dredging of the natural Kootenay Lake outlet upstream of Corra Linn Dam to reduce outlet hydraulic control (1930s), an international treaty signed to increase winter levels and decrease spring levels in Kootenay Lake, impacting natural backwater proﬁles to Bonners Ferry, ID (1938), and completion of Libby Dam (1973) (dates of events from Tetratech, 2004). The key challenge to determine the Libby reservoir operations-based losses to the Lower Kootenai is to isolate the reservoir’s impacts from the confounding effects of the basin-management actions described above. 4.1.2.

Motivation for assessment

Libby Dam and reservoir were completed in 1974. Several adverse ecosystem impacts across the river ﬂoodplain continuum have been attributed to the facility. These impacts include negligible recruitment of native ﬁsh such as white sturgeon (Acipenser transmontanus) and burbot (Lota lota) (Hoffman et al., 2002). Also attributed to reservoir operations is a disruption in recruitment of black cottonwood (Populus trichocarpa), a key structural element in riparian forests in the region (Polzin and Rood, 2000). As the native ecosystem served as the basis for the subsistence economy of the Kootenai people historically, the fate of the species described above and that of other species are directly linked with the long-term sustainability of this indigenous culture. Therefore, an assessment of ecosystem losses due to Libby Dam and Reservoir operations is currently being led by the Kootenai Tribe of Idaho with the cooperation of several government and non-government entities. The study presented here is a part of this larger effort. 4.2.

Conceptual model application to lower Kootenai

The conceptual model as shown in Fig. 23.3 has been developed for the lower Kootenai ﬂoodplain.

K. Jorde et al.

622 4.2.1.

Scenarios

Based on an understanding of the relative magnitudes of hydrodynamic impacts attributable to the management actions described above, reference scenarios have been established that represent historic, contemporary pre-dam and post-dam eras. Comparison between historic and post-dam scenarios is used to quantify total perturbation of the system, incorporating land-use impacts, downstream hydraulic impacts, and Libby Dam impacts. Comparison between pre-dam and post-dam scenarios allows isolation of the impacts of Libby Dam and Reservoir (Burke, 2006). The historic scenario is representative of the condition prior to 1938. This date is signiﬁcant as it marks the International Joint Commission (IJC) Order of Approval, which modiﬁed the downstream boundary condition due to Kootenay Lake. The change in the downstream boundary condition represents the ﬁrst large magnitude change to Lower Kootenai River function (Burke, 2006). The IJC Order dictated manipulation of Kootenay Lake levels through operation of Corra Linn Dam, resulting in increased winter and decreased spring lake levels for ﬂood control purposes (IJC, 1938). The contemporary pre-dam scenario is representative of the period 1938–1967. This period is characterized by extensive levee construction, construction of an inline run-of-river hydropower plant on the Moyie River (1949), a larger dam on the Duncan River (1967 – tributary to the North Arm of Kootenay Lake), and other land-use-based impacts. At the end of this period, impacts to Kootenai River function due to land use and basin-management actions (with the exception of Libby Dam construction) are in place (Burke and Jorde, 2004). The post-dam scenario represents the condition following completion of Libby Dam (1975 – present). The period from 1967 (commencement of Libby construction) to 1974 (full pool achieved) has been omitted due to variable conditions during facility construction and reservoir ﬁlling. Compared to the end of contemporary pre-dam scenario, the single incremental intervention during the post-dam period is the construction and operation of Libby Dam and reservoir (Burke and Jorde, 2004). The sequence of land use and basin-management actions, together with delineation of the reference scenarios described above, is represented in Fig. 23.3. Starting in 1993, Bonneville Power Administration has provided experimental ﬂow releases from Libby Dam to stimulate spawning by white sturgeon, with typical discharges during the months of May and June increased by approximately 140 m3/s. This adjustment in operations is important for its intent to mimic the timing of the historic spring snowmelt peak. However, due to limitations in powerhouse and spill capacities, and ﬂood control requirements, the maximum experimental ﬂows are still less than 50% of historic median annual peak ﬂows (Dibrani, 2003). Fig. 23.4 shows the relationship between historic and contemporary (including experimental ﬂows) mean ﬂows for the USGS gage 12305000, Kootenai River at Leonia (Burke and Jorde, 2004). 4.2.2.

Summary of data available to support evaluation of reference scenarios

Flow data from nine mainstem gages (Fig. 23.2) and relevant tributaries were obtained from USGS and Water Survey of Canada (WSC) databases. Correlations

Reservoir operations, physical processes, and ecosystem losses

623

1600

Mean 29-70 1400

Mean 72-92 Mean 93-01

Discharge (m³/s)

1200 1000 800 600 400 200

p Se

Au g

ly Ju

ai

ne Ju

M

ril Ap

ch

b

ar M

Fe

n Ja

ec D

v No

O

ct

0

Figure 23.4. Timing and magnitude for Kootenai River ﬂows at USGS gage No. 12305000, Kootenai River at Leonia, adjacent to Idaho-Montana state boundary for three different time periods.

between gages were used to create a complete ﬂow record from 1910 to 2003 for US gages and 1914–2003 for Canadian gages. Repeat surveys of the lower river ﬂoodplain topography and channel bathymetry dated 1929 (USGS) and 1956 (USACE) are used in evaluation of historic and contemporary pre-dam scenarios, respectively. Contemporary scenario ﬂoodplain topography was captured through LIDAR survey of the lower ﬂoodplain (Geoengineers, 2005), while the contemporary channel bathymetry is represented by data collected by the USGS (Barton et al., 2004) and others (Dibrani, 2003; Zelch, 2003).

4.2.3.

Operational elements

In order to assess the degree of hydrologic alteration caused by the facility, the longterm time series for seven mainstem gages were evaluated using the indicators of hydrologic alterations (IHA) method (Richter et al., 1996, 1997, 1998) and software (Nature Conservancy, 2003). In their 2003 analysis, Olden and Poff (2003) found that the suite of indices resulting from the IHA method adequately characterize the principal components of ﬂow regimes. When pre-dam and post-dam periods are analyzed, the subsequent hydrologic alteration due to the dam’s operations can be evaluated (Richter et al., 1996, 1997). By conducting this analysis for multiple gages throughout a drainage basin, the spatial distribution of the alteration can be assessed (Richter et al., 1998). Changes of the hydrologic regime due to land-use or climatic change are negligible compared to the dam operation (Burke, 2006).

K. Jorde et al.

624 4.2.4.

Hydraulic simulations

The underlying physical processes used for the case study applications are based on results from one-dimensional hydrodynamic simulation models, which solve the St.-Venant equations for unsteady non-uniform ﬂow. For the simulation the software tool MIKE 11 from Danish Hydraulic Institute (www.dhigroup.com) has been used. In total, three different MIKE 11 models have been developed, one for each condition, historic, contemporary pre-dam and post-dam. River-bed bathymetry and ﬂoodplain topography was based on survey datasets mentioned above. For example, the contemporary model is based on 245 cross-sections, spaced between 100 and 2 km. Upstream of Bonner’s Ferry no historical cross-section surveys were available and therefore the contemporary cross-sectional data were also used in the historic and pre-dam models. Evaluations of post-dam channel change in the canyon reach suggest that this is a valid assumption (Burke, 2006). Upstream boundary conditions (hydrologic data) and downstream boundary conditions (water-level elevations) and data for model calibration were available mostly from USGS gaging stations. Once calibrated, the models were run for the years selected to represent certain scenarios. Results from MIKE 11 include information on the river channel, such as local waterlevel elevations, water depths, cross-section averaged ﬂow velocities and shear stress, stream power, each of them as time series with 1-day time steps. The digital elevation models (DEM) were based on historical survey maps that were digitized and, for the contemporary ﬂoodplain based on 2004 LIDAR data. Time series of water-level elevations generated by MIKE 11 were combined with the DEM of the ﬂoodplain in a GIS environment, thus generating time series of local water depths or water-level elevations for different scenarios. 4.2.5.

Case study example applications

We use comparisons between historic, contemporary pre-dam and post-dam reference scenarios to exhibit operational impacts on fundamental ecosystem processes, which are dependent on different characteristics of the hydrologic regime. First, we discuss ﬁrst-order impacts on the hydrologic regime and on sediment supply. We then discuss two examples of higher order impacts. Channel bed mobility is an example of a second order impact, and cottonwood recruitment is an example of a third order impact. 4.2.6.

Hydrologic regime

In order to assess the degree of hydrologic alteration caused by the facility, the longterm time series for seven mainstem gages were evaluated using the IHA methodology (Richter et al., 1996, 1997, 1998) and software. The IHA analysis suggested that the most signiﬁcant aspects of hydrologic alteration include reduced maximum ﬂows during the timing of the historic snowmelt peak, increased minimum ﬂows during the pre-regulation winter low-ﬂow period, and reduced constancy of the historically highly predictable annual ﬂow pattern. These results suggest that the instream ﬂow parameters and ecological processes dependent on peak and low-ﬂow

Reservoir operations, physical processes, and ecosystem losses

625

magnitudes and timing, and sustained hydrograph trends, are most likely inﬂuenced by Libby Dam operations (Burke, 2006). 4.2.7.

Sediment supply

While not evaluated explicitly based on ﬁeld studies, existing information resources were evaluated to determine ﬁrst-order impacts of Libby Dam operations on sediment supply. The studies and data suggest that supply of both suspended sediment and bed material has been signiﬁcantly curtailed by the presence and operation of Libby Dam. Based on ﬁeld studies by others, it appears that the reduction in supply of ﬁner sediment is more pronounced than the reduction of bed sediment in the canyon reach. The river appears to be capable of removing ﬁner materials from the channel margins. In the absence of replenishment from upstream sources, a deﬁcit of ﬁne-grained materials along the channel margins in the canyon has been documented, a phenomenon which corresponds with other regulated rivers (Williams and Wolman, 1984). The river is less capable of mobilizing the coarse bed material due to reduced ﬂow magnitudes. The reduction of supply in bed material has been offset by reduced capacity to mobilize bed material resulting in a stabilized channel condition (Burke, 2006). The braided reach is located within a major depositional zone, and has been active in the post-dam period, though within historical bounds (Tetratech, 2004). Quantitative data linking upstream sediment supply to channel changes in this reach are sparse, limiting the determination of direct impacts from Libby Dam operations (Burke, 2006). Comparison of repeat data sources for the meandering reach suggests that areas of deposition have developed and shifts in the bed proﬁle may have occurred as a result of Libby Dam operation. Deposition has likely occurred as a result of reduced sediment transport capacity during the post-dam period. The deposition may also be correlated with the selective removal of ﬁne sediment from upstream reaches as a result of regulation. The abundance of ﬁne sediment in mobile bed forms in the white sturgeon spawning reach is the subject of ongoing research (e.g., Barton, 2004; Barton et al., 2005; Berenbrock, 2005; Berenbrock and Bennett, 2005). 4.2.8. Native cottonwood seedling recruitment – magnitude, timing, and rate of change The importance of cottonwood trees (Populus spp.) as a structural element in riparian areas of western North America is widely acknowledged (Braatne et al., 1996; Rood et al., 2005). Limited recruitment of Black Cottonwood in the Lower Kootenai River since the closure of Libby Dam has been primarily attributed to the virtual elimination of upstream sediment supply and selective removal of ﬁner sediments from potential recruitment sites, and modiﬁcations of the downstream hydrologic regime (Polzin and Rood, 2000). The streamﬂow pattern requirements for successful seedling recruitment have been detailed in several recent studies, often summarized in the ‘Cottonwood Recruitment Box Model’ or CRBM (Mahoney and Rood, 1998; Amlin and Rood, 2002; Rood et al., 2005). These patterns are consistent with the natural ﬂow pattern of northern

626

K. Jorde et al.

snowmelt-dominated river systems. The general components of the CRBM include the following: (1) a large late spring peak or ‘spring freshet’ to mobilize and redistribute sediments, followed by (2) gradual recession to baseﬂow during and following the summer seed release period, and (3) low ﬂows following the recruitment period to prevent mortality by scouring newly established seedlings (Mahoney and Rood, 1998). These criteria can be evaluated from the hydraulic analysis described earlier. The components of the CRBM combine to describe the process of cottonwood recruitment. Mobilization of sediments from the peak ﬂows provides barren recruitment sites resulting from fresh scour or deposition, and the gradual rate of recession allows newly established seedling roots to stay in contact with adequate soil moisture as they elongate (Mahoney and Rood, 1998). Laboratory studies have shown optimal rates for water table decline are approximately 2.5 cm/day, which could be transferred as a criteria for stage decline to natural settings with well-drained gravel substrate materials. Various species of Populus are able to survive a greater range of stage recession (Amlin and Rood, 2002). In their evaluations of stage decline rates on other rivers, Rood and Mahoney (2000) and Braatne et al. (2007) use a convention where a 3-day moving average of daily stage decline from 0 to 5 cm/day is considered favorable for seedling recruitment, while rates between 5 and 10 cm/day are considered stressful, and other rates are considered lethal. As a surrogate for linear application of the 2.5 cm/day rate on a daily basis, this convention is considered to capture the essence of the required stage decline trend, while allowing for inherent variability in ﬁeld settings. Braatne et al. (2007) further utilize the concept of a ‘mortality coefﬁcient’ in their evaluation, which is a weighting convention that allows a certain proportion of ‘lethal’ days to occur during the stage recession period. This convention also accounts for natural variability during the stage decline period, suggesting that the capillary fringe followed by the elongating roots is not desiccated by limited occurrence of lethal rates. Finally, observations have shown that naturally recruited cottonwoods occur within consistent elevation ranges relative to late summer baseﬂow (Mahoney and Rood, 1998). This elevation band has been shown to typically range between 0.5 and 1.5 m above baseﬂow for many rivers, though may have upper bound as great as 3 m above baseﬂow or greater on large rivers with high snowmelt runoff volumes (Burke, 2006). This elevation window is considered to be the potential ‘recruitment band’ in the recruitment box model. For each year, the correct ﬂow attributes may converge over a subsection of this band, resulting in a unique pattern of seedling establishment during those years where recruitment occurs. In support of the Kootenai River Assessment, Burke (2006) evaluated riparian recruitment potential within the bed and banks of the active river channel in the study reach for historic, pre-dam and post-dam periods. He found that modiﬁed Kootenay Lake levels (late summer drafting) resulting from the 1938 IJC Order of Approval inhibited cottonwood recruitment potential in the meandering reach for the pre-dam scenario by causing sustained stage decline rates in excess of documented tolerances for cottonwood seedlings beyond the end of the annual seed dispersion period. Following closure of Libby Dam, cottonwood recruitment was shown to be further curtailed in the meandering reach and highly disrupted in the

Reservoir operations, physical processes, and ecosystem losses

627

braided and canyon reaches due to signiﬁcant hydrologic alteration including timing and rates of changes. Interestingly however, following initiation of experimental ﬂow releases in the early 1990s to spur recovery of native ﬁsh species, conditions for recruitment of cottonwoods appear to have improved in some relatively wet years. Fig. 23.5 shows a representation of the spatial distribution of riparian recruitment potential developed by combining the results of the hydrodynamic model with terrain data and applying the CRBM as described above. The location shown in Fig. 23.5 is centered at approximate river km 251 within the braided reach. The top two panes of Fig. 23.5 show time series of the water-level elevations during the recruitment and growing season of the years of 1944 and 2001 and the principle of application of the recruitment box criteria to this pre-dam/post-dam pair of years for this location. The lower panes of Fig. 23.5 show composites of potential recruitment patches that result when the recruitment bands corresponding to the series of predam (left-side) and post-dam (right-side) years are projected over the example reach topography. The pane that represents the pre-dam years shows that recruitment potential was indicated for only 2 of the 4 years analyzed, the dry years (1944 and 1945), resulting in a variety of recruitment patches over a varied range of elevations and spatial extents in the reach. The results also show that for the average year (1955) and the wet year (1950) no recruitment was indicated even before the dam at this location. The right lower panel of Fig. 23.5 shows that recruitment potential for all paired post-dam years is zero at this location. Comparison of composite images such as these for pre-dam and post-dam periods or other scenarios allows direct comparison of the regulation or other impacts on riparian habitat development. 4.2.9.

Channel bed mobility: magnitude and timing

Periodic channel bed movement and reorganization has been shown to be relevant for a number of ecologically signiﬁcant processes, including maintenance of ﬁsh spawning materials, development and maintenance of diverse aquatic habitat conditions, and riparian succession (Mahoney and Rood, 1998). To evaluate the general impact of reservoir operations on the spatial and temporal pattern of bed mobility as part of the Kootenai assessment, Burke (2006) conducted incipient motion-based calculations spaced over a 90.8 km river reach (river km 249.7–340.5). Little quantitative evidence of the pre-dam bed composition is available. Based on disrupted upstream sediment supply, an operational regime characterized by relatively lower yet much more frequent peak ﬂows (hydropeaking), and anecdotal evidence (Polzin and Rood, 2000), it was assumed that the pre-dam channel bed composition was at a minimum as ﬁne as the contemporary channel bed, and was likely ﬁner in composition. Then, the pre-dam and post-dam ﬂows were applied to the post-dam D50 bed surface sizes to assess the relative ability of the pre-dam and post-dam ﬂow patterns to move the channel bed. Based on evidence of channel armoring, it was further assumed that the surface sediments were limiting in evaluation of bed mobility (Burke, 2006). The evaluation utilized provisional bed surface samples collected mid-channel in water depths ranging from approximately 2–10 m with a vessel-mounted underwater video camera. The captured substrate images were converted via video granulometry

628

Composite-1944, 1945, 1950, 1955

1978, 1979, 1999, 2001: no potential recruitment

1950, 1955: no potential recruitment Figure 23.5. Water-level elevations (top panels) and potential for cottonwood recruitment (lower panels) in paired pre- and post-dam years. The left column represents pre-dam conditions, the right column represents post-dam conditions. The paired years are 1944/2001 (dry years, also shown in top panels), 1945/1979 (dry), 1955/1978 (average), 1950/1999 (wet). The 2.5 cm/day stage decline line is only as general reference.

K. Jorde et al.

1944 – potential recruitment band 1945 – potential recruitment band

Composite-1978, 1979, 1999, 2001

Reservoir operations, physical processes, and ecosystem losses

629

software to area-by-weight (photo-sieved) samples, which were subsequently converted to volume-by-weight samples using methods such as described in Bunte and Abt (2001). There is inherent uncertainty in selection of a critical Shields parameter (tc50 ), with values applicable to gravel-bedded rivers cited in the literature ranging from 0.03 to 0.1 (Bufﬁngton and Montgomery, 1997). Since the objective of this analysis was to demonstrate relative trends in bed mobility, a conservative value (0.03) was selected for the analysis. With this conservative value, relatively lower critical shear stresses are predicted for a given D50 grain size, indicating more mobility across all scenarios considered than would be obtained using higher Shields values. For this evaluation, 1-D hydrodynamic simulations were run over the water year (October 1–September 30) for paired pre-dam (1956) and postdam (1996) years having relatively high annual peak ﬂow and mean annual ﬂow (Burke, 2006). The ‘‘signature of bed mobility’’ for both of these years is represented in Fig. 23.6. For reference, also shown on the ﬁgure is the thalweg proﬁle over the reach, with major tributary conﬂuences indicated. The pre-dam high-ﬂow result (1956) shows a regular pattern of mobility distributed over the study reach, with mobility occurring exclusively during the April–July high-ﬂow period. The pattern of pre-dam mobility uniformly precedes the cottonwood and willow recruitment period (timing of seed dispersal and recruitment period are indicated in Fig. 23.6) described by the ‘riparian recruitment box model’ which summarizes the documented ﬂow-pattern dependency for these trees (Mahoney and Rood, 1998; Amlin and Rood, 2002). Mobility that directly precedes the timing of seed dispersal should result in fresh recruitment surfaces for the riparian trees. For the post-dam high ﬂow year (1996), signiﬁcant mobility is also indicated. Examination of this mobility pattern shows that ﬂows are maintained at higher levels throughout the year, though the mobility is limited to fewer sites. Sustained mobility through fall and winter may have the adverse effect of scouring successfully recruited seedlings from the preceding recruitment period. The fact that mobility is less evenly distributed over the study reach, but for longer periods suggests that while the locations with larger grain sizes are not being scoured, the ﬁner grained locations are being scoured more frequently. This is consistent with the observed deﬁcit of ﬁner grained materials in the study reach since regulation commenced (Polzin and Rood, 2000; Burke, 2006).

4.3.

Discussion of case study application

The results shown here are a ﬁrst attempt to quantify a speciﬁc component of ecosystem functional losses caused by reservoir operation. Regarding the study on cottonwood recruitment potential the loss shown here is complete, or 100%, as compared to pre-dam conditions. The results show that historically different hydrologic years provide recruitment chances for cottonwoods at different elevations along the channel, or patches. Thus, a balanced age class distribution among cottonwoods in a pristine environment is a consequence of interannual alterations of the hydrological regime.

K. Jorde et al.

630

660

620

Bonners Ferry

Moyie River

Yaak River sediment motion seed dispersal and recruitment period

580 560

520

258.2 255.6 252.4 249.7

264.1

272.2

282

286.7

540

294.5

306.1 302.8 299.6

1956

333.9 330.4 328.5 325 322.4 320 318.2

river kilometer

600

elevation (m-NAVD 88)

640

340.5

9/30 9/16 9/2 8/19 8/5 7/22 7/8 6/24 6/10 5/27 5/13 4/29 4/15 4/1 3/18 3/4 2/18 2/4 1/21 1/7 12/24 12/10 11/26 11/12 10/29 10/15 10/1

Fisher River

water year date

Libby Dam

660

620

Bonners Ferry

580

560

540

258.2 255.6 252.4 249.7

264.1

272.2

282

294.5

1996 306.1 302.8 299.6

333.9 330.4 328.5 325 322.4 320 318.2

340.5

sediment motion seed dispersal and recruitment period

Moyie River

Yaak River

600

elevation (m-NAVD 88)

640

286.7

9/30 9/16 9/2 8/19 8/5 7/22 7/8 6/24 6/10 5/27 5/13 4/29 4/15 4/1 3/18 3/4 2/18 2/4 1/21 1/7 12/24 12/10 11/26 11/12 10/29 10/15 10/1 river kilometer

Fisher River

water year date

Libby Dam

520

Figure 23.6. ‘‘Signature of bed mobility’’ for two paired years representing the pre- and post-dam situation. The thalweg line refers to the elevation on the right vertical axis. The left vertical axis represents time and the symbols for sediment motion indicate the time and location throughout the river length and year modeled where sediment at that location is mobilized.

For a complete picture, more water years and potential recruitment locations are presently analyzed. The demonstrated alterations of water-level ﬂuctuations and sediment transport in comparable hydrologic years do in combination not only affect cottonwood recruitment but also, alone or jointly with other physical processes, other types of ﬂoodplain vegetation or gravel spawning ﬁsh. At this level, the simulation results are in agreement with ﬁeld observations (no recruitment), however

Reservoir operations, physical processes, and ecosystem losses

631

more validation is needed. In the future the model will be applied to the entire ﬂoodplain, also including historic ﬂoodplain topography and historic river crosssections without dikes. The model will also be used to develop instream ﬂow releases from Libby Dam and other strategies, such as levee removal, to support cottonwood recruitment in the future.

5.

Discussion and conclusions

As we understand more about the complexity of riverine ecosystems and the underlying physical processes it appears more relevant to leave the natural ﬂow regime untouched to maintain those processes and consequent ecological functions. The reality, however, is that most of our rivers are heavily impacted by dams and reservoirs and that there is little room for true restoration efforts. Also, in many countries, engineers keep building and operating dams the same ways they did half a century ago claiming there is no ecological damage. Even where the damage has happened and is understood, there is little willingness to do anything about it as water and energy get scarcer and more valuable. The most promising way to bring ecologists and dam owners together is by searching for win-win situations and by maximizing the beneﬁts that can be achieved for the ecology with a given quantity of water or energy. It is therefore crucial to analyze quantitatively where and which ecosystem functions have been lost. Since ecosystem functions cannot be reestablished without providing the underlying physical processes, it is necessary to understand these, identify the essential components that support certain ecosystem functions, and investigate how these processes can be brought back – at least partially. The study we presented here is a ﬁrst attempt to quantify the operational losses of cottonwood stands, an essential component of North American riverine ecosystems. The development of the conceptual model helps understand which alterations in the ﬂoodplain other than the dam operation might contribute to ecosystem functional losses. By setting up scenarios for numerical modeling of relevant processes and the differential evaluation of the results for both, physical processes and ecosystem function, we can identify the amount of losses attributable to the dam operation or to other factors, respectively. In the same way, these tools can then be used to investigate different restoration strategies ranging from changing dam operation to removing dikes or opening ﬂoodways in the ﬂoodplain. The goal is to change dam operation in such a way that the losses in water or energy are as small as possible and help in achieving the maximum ecological beneﬁt. For cottonwood recruitment, this could mean that in years where surplus water is available, a certain ﬂood regime is released downstream which creates the required conditions over certain areas. Even if cottonwood recruitment was only supported once every 5–10 years, the situation would be much better than the existing one. Unfortunately, property developers often construct dwellings close to the river channel as soon as a river course is regulated by a dam (and therefore apparently ‘‘safe’’), expensive homes are erected on gravel bars and ﬂooding becomes unthinkable.

632

K. Jorde et al.

The modeling efforts shown here require good data regarding hydrological data, river cross-sections, and ﬂoodplain topography, both historic and contemporary. In the case of the Kootenai River, this information is available, in other cases this may not be so. Remote sensing technologies such as LIDAR or hyperspectral imagery will become increasingly available and more affordable in the future. Within the general framework of costs of building, operating, and maintaining large dams, beneﬁts provided and environmental damage caused by dams, acquiring remote sensing data sets is not a signiﬁcant cost factor. The ability to fund such efforts depends on the willingness of stakeholders and decision makers to address operational losses caused by existing or future dams adequately or not. A more difﬁcult question is the reconstruction of historical conditions. While in most cases photographic or anecdotal evidence exists that describes how a ﬂoodplain was looking before dams were put in place, there will often be a lack of accurate physical data. In some cases pre-dam river cross-sections will be available. Information on historic ﬂoodplain topographies could possibly be derived from ground data collection (digging of trenches, etc.) in very small scales or from ground penetrating radar data in larger scales. It depends on the objective of a speciﬁc study if quantitative modeling of the historic situation is required or not. If restoration strategies are to be evaluated, modeling of the historic situation is not required necessarily. We plan to analyze the entire Lower Kootenai Floodplain over the next few years and to develop more numerical tools that follow the conceptual model. The losses of different types of ﬂoodplain and related habitats that were historically abundant and are now largely missing will be quantiﬁed with the help of numerical tools. These tools will also be used to address restoration strategies.

Acknowledgements This work is funded by the Bonneville Power Administration through the Kootenai Tribe of Idaho. The Boise ofﬁce of the US Geological Survey provided river cross-section surveys. The Danish Hydraulic Institute granted access to their hydrodynamic modeling software MIKE 11 for use in this study.

References Akeret, E., 1982. Schlussbereicht der Interdepartementalen Arbeitsgruppe Restwasser. Eidgeno¨ssisches Verkehrs-und Energiewirtschaftsdepartement. Bern, Switzerland. Amlin, N.M., Rood, S.B., 2002. Comparative tolerances of riparian willows and cottonwoods to watertable decline. Wetlands 22, 338–364. Amoros, C., Bornette, G., 2002. Connectivity and biocomplexity in waterbodies of riverine ﬂoodplains. Freshw. Biol. 47, 761–776. Barton, G.J., 2004. Characterization of channel substrate, and changes in suspended-sediment transport and channel geometry in White Sturgeon Spawning Habitat in the Kootenai River near Bonners Ferry, Idaho, following the closure of Libby Dam. U.S. Geological Survey.

Reservoir operations, physical processes, and ecosystem losses

633

Barton, G.J., McDonald, R.R., Nelson, J.M., Dinehart, R.L., 2005. Simulation of ﬂow and sediment mobility using a Multidimensional Flow Model for the White Stugeon Critical-Habitat Reach, Kootenai River near Bonners Ferry, Idaho. Barton, G.J., Moran, E.H., Berenbrock, C., 2004. Surveying Cross Sections of the Kootenai River between Libby Dam, Montana and Kootenay Lake, British Columbia, Canada. USGS, Boise, ID. Berenbrock, C., 2005. Simulation of Hydraulic Characteristics in the White Sturgeon Spawning Habitat of the Kootenai River Near Bonners Ferry, Idaho. U.S. Geological Survey, Boise, ID. Berenbrock, C., Bennett, J.P., 2005. Simulation of Flow and Sediment Transport in the white Sturgeon Spawning Habitat of the Kootenai River Near Bonners Ferry, Idaho. U.S. Geological Survey, Boise, ID. Boon, P.J., 1992. Essential elements in the case for river conservation. In: Boon, P.J., Calow, P., and Petts, G. (Eds), River Conservation and Management. Wiley, Chichester, pp. 11–33. Bovee, K.D., 1986. Development and evaluation of habitat suitability criteria for use in the instream ﬂow incremental methodology. US Fish and Wildlife Service, Instream Flow Information Paper No. 21, Biologic Report 86 (7), 235pp. Braatne, J.H., Jamieson, B., Gill, K., Rood, S.B., 2007. Instream ﬂows and declines of riparian cottonwood along the Yakima River, Washington, USA. Riv. Res. Appl. 23(3), 247–267. Braatne, J.H., Rood, S.B., Heilman, P.E., 1996. Life history, ecology and conservation of riparian cottonwoods in North America. In: Stettler, R.F., Bradshaw, H.D., Heilman, P.E., and Hinckley, T.M. (Eds), Biology of Populus and its Implication for Management and Conservation. NRC Press, Ottawa, ON, Canada. Bratrich, C., Truffer, B., 2001. Green electricity certiﬁcation for hydropower plants – concept, procedure, criteria. Green Power Publications, Issue 7. EAWAG, Kastanienbaum, Switzerland, 123pp. Bufﬁngton, J.M., Montgomery, D.R., 1997. A systematic analysis of eight decades of incipient motion studies, with special reference to gravel-bedded rivers. Water Resour. Res. 33, 1993–2029. Bufﬁngton, J.M., Montgomery, D.R., 1999. Effects of sediment supply on surface textures of gravel-bed rivers. Water Resour. Res. 35, 3523–3530. Bunte, K., Abt, S.R., 2001. Sampling surface and subsurface particle-size distributions in wadable graveland cobble-bed streams for analysis in sediment transport, hydraulics, and streambed monitoring. In: F. S. U.S. Department of Agriculture, Rocky Mountain Research Station (Ed.), Gen. Tech. Rep. RMRS-GTR-74. U.S. Department of Agriculture, Forest Service, Rocky Mountain Research Station, Fort Collins, CO, USA, 428pp. Burke, M. and Jorde, K., 2004. Physical process based conceptual framework for assessment of ecosystem losses in river ﬂoodplain systems due to reservoir operations. In: Garcia, D. (Ed.), Fifth International Symposium on Ecohydraulics. IAHR Congress Proceedings, Madrid, Spain, pp. 840–844. Burke, M.P., 2006. Linking hydropower operation to modiﬁed ﬂuvial processes downstream of Libby Dam, Kootenai River, USA and Canada. College of Graduate Studies, Department of Civil Engineering. University of Idaho, Moscow, ID. Church, M., 1995. Geomorphic response to river ﬂow regulation: case studies and time-scales. Regul. Rivers Res. Manage. 11, 3–11. Church, M., 2002. Geomorphic thresholds in riverine landscapes. Freshw. Biol. 47, 541–557. Connell, J.H., 1978. Diversity in tropical rain forests and coral reefs. Science 199, 1302–1310. COST-626, 2005. European Cooperation in the ﬁeld of Scientiﬁc and Technical Research (COST) 626, European Aquatic Habitat Modelling Network. In: Harby, A. (Ed.), European Aquatic Habitat Modelling Network. National Environmental Research Institute, Denmark, Silkeborg, Denmark. Craig, J.F., 1999. Large dams and freshwater ﬁsh biodiversity. In: The World Commission on Dams (Ed.), Dams and Development: A New Framework for Decision-Making. Earthscan Publications, Virginia, USA. Crowder, D.W., Diplas, P., 2000. Evaluating spatially explicit metrics of stream energy gradients using hydrodynamic model simulations. Can. J. Fish. Aquat. Sci. 57, 1497–1507. Dibrani, B., 2003. Simulation of Flood Induced Removal of Alluvial Fans from Tributaries of the Kootenai River. Department of Civil Engineering, Universita¨t Stuttgart, Stuttgart. Egger, G., Angermann, K., Kerle, F., et al., 2005. RiverSmart: a decision support system for restoration planning. In: Harby, A. (Ed.), COST 626 – European Aquatic Habitat Modeling Network. National Environmental Research Institute, Silkeborg, Denmark, pp. 57–64.

634

K. Jorde et al.

Friedman, J.M., Auble, G.T., 1999. Mortality of riparian box elder from sediment mobilization and extended inundation. Regul. Rivers Res. Manage. 15, 463–476. Frisell, C.A., Liss, W.J., Warren, C.E., Hurley, M.D., 1986. A hierarchical framework for stream habitat classiﬁcation: viewing streams in a watershed context. Environ. Manage. 10, 199–214. Geoengineers, 2005. LIDAR Survey of the Lower Kootenai Floodplain, Spokane, WA. Gouraud, V., Sabaton, C., Capra, H., 2004. Role of habitat variability in trout population dynamics: application of a dynamic population model to three French rivers. Hydroecol. Appl. 1-2004, 221–244. Graf, W.L., 1999. Dam nation: a geographic census of American dams and their large-scale hydrologic impacts. Water Resour. Res. 35, 1305–1311. Hauer, F.R., Smith, R.D., 1998. The hydrogeomorphic approach to functional assessment of riparian wetlands: evaluating impacts and mitigation on river ﬂoodplains in the U.S.A. Freshw. Biol. 40, 517–530. Heiler, G., Hein, T., Schiemer, F., 1995. Hydrological connectivity and ﬂood pulses as the central aspects for the integrity of a river-ﬂoodplain system. Regul. Rivers Res. Manage. 11, 351–361. Hill, B.H., Herlihy, A.T., Kaufmann, P.R., et al., 2003. Assessment of streams of the eastern United States using a periphyton index of biotic integrity. Ecol. Ind. 2, 325–338. Hoffman, G., Skaar, D., Dalbey, S., et al., 2002. Instream ﬂows incremental method for Kootenai River. Bonneville Power Administration, 86pp. electronic pages. Huggenberger, P., Hoehn, E., Beschta, R., Woessner, W., 1998. Abiotic aspects of channels and ﬂoodplains in riparian ecology. Freshw. Biol. 40, 407–425. IJC, 1938. In the matter of the application of the West Kootenay Power and Light Company, Limited, for permission to construct and operate certain works in and adjacent to the channel of the Kootenay River in the Province of British Columbia, and for the right to store water in Kootenay Lake in the said province of British Columbia. International Joint Commission, New York, USA. Johnson, N., Revenga, C., Echeverria, J., 2001. Managing water for people and nature. Science 292, 1071–1072. Jorde, K., Bratrich, C., 1998. Inﬂuence of river bed morphology and ﬂow regulations in diverted streams: effects on bottom shear stress patterns and hydraulic habitat. In: Bretschko, G. and Helesic, J. (Eds), Advances in River Bottom Ecology IV. Backhuys Publishers, Leiden, The Netherlands, pp. 47–63. Junk, W.J., Baylay, P.B., and Sparks, R.E., 1989. The ﬂood pulse concept in river ﬂoodplain systems. In: Dodge, D.P. (Ed.), Proceedings of the International Large River Symposium, Can. Spec. Publ. Fish. Aquat. Sci. Vol. 106, pp. 110–127. Kootenai Tribe of Idaho and Montana Fish Wildlife & Parks. 2004. Kootenai Subbasin Plan. Part I: Kootenai River Subbasin Assessment. A report prepared for the Northwest Power and Conservation Council, Portland, OR. Lamouroux, N., Capra, H., Poully, M., 1998. Predicting habitat suitably for lotic ﬁsh: linking statistical hydraulic models with multivariate habitat use models. Regul. Riv. Res. Manage. 14(1), 1–11. Leclerc, M., Saint-Hilaire, A., Bechara, J., 2003. State-of-the-art and perspectives of habitat modelling for determining conservation ﬂows. Can. Water Resour. J. 28, 135–151. Ligon, F.K., Dietrich, W.E., Trush, W.J., 1995. Downstream ecological effects of dams. Bioscience 45, 183–192. Mahoney, J.M., Rood, S.B., 1998. Streamﬂow requirements for cottonwood seedling recruitment: an integrative model. Wetlands 18, 634–645. Malard, F., Uehlinger, U., Zah, R., Tockner, K., 2006. Flood-pulse and riverscape dynamics in a braided glacial river. Ecology 87, 704–716. McCully, P., 1996. Silenced Rivers – The Ecology and Politics of Large Dams. Zed Books, London. Merritt, D.M., Cooper, D.J., 2000. Riparian vegetation and channel change in response to river regulation: a comparative study of regulated and unregulated streams in the Green River Basin, USA. Regul. Rivers Res. Manage. 16, 543–564. Montgomery, D.R., Bufﬁngton, J.B., 1998. Channel processes, classiﬁcation, and response. In: Naiman, R. and Bilby, R. (Eds), River Ecology and Management. Springer Verlag, New York. Montgomery, D.R., Bufﬁngton, J.M., 1997. Channel-reach morphology in mountain drainage basins. Geol. Soc. Am. Bull. 109, 596–611.

Reservoir operations, physical processes, and ecosystem losses

635

Naiman, R.J., Bilby, R.E., Bisson, P.A., 2000a. Riparian ecology and management in the Paciﬁc Coastal rain forest. Bioscience 50, 996–1011. Naiman, R.J., Elliot, S.R., Helﬁeld, J.M., O’Keefe, T.C., 2000b. Biophysical interactions and the structure and dynamics of riverine ecosystems: the importance of biotic feedbacks. Hydrobiologica 410, 79–86. Nature_Conservancy, 2003. Indicators of hydrologic alteration (IHA): software for understanding hydrologic changes in ecologically-relevant terms. http://www.freshwaters.org/tools/ Olden, J.D., Poff, N.L., 2003. Redundancy and the choice of hydrologic indices for characterizing streamﬂow regimes. River Res. Appl. 19, 101–121. Petts, G., 1984. Impounded Rivers: Perspective for Ecological Management. Wiley and Sons, Chichester. Poff, N.L., Allan, J.D., Bain, M.B., et al., 1997. The natural ﬂow regime. Bioscience 47, 769–784. Polzin, M.L., Rood, S.B., 2000. Effects of damming and ﬂow stabilization on riparian processes and black cottonwoods along the Kootenay River. Rivers 7, 221–232. Poole, G.C., Stanford, J.A., Frissell, C.A., Running, S.W., 2002. Three-dimensional mapping of geomorphic controls on ﬂood-plain hydrology and connectivity from aerial photos. Geomorphology 48, 329–347. Pringle, C., 2003. What is hydrologic connectivity and why is it ecologically important? Hydrol. Process. 17, 2685–2689. Revenga, C., Brunner, J., Henninger, N., et al., 2000. Pilot Analysis of Global Ecosystems: Freshwater Systems. In: World Resources Institute (Ed.), World Resources Institute, Washington, DC, 65pp. Richards, K., Brasington, J., Hughes, F., 2002. Geomorphic dynamics of ﬂoodplains: ecological implications and a potential modelling strategy. Freshw. Biol. 47, 559–579. Richter, B.D., Baumgartner, J.F., Braun, D.P., Powell, J., 1998. A spatial assessment of hydrologic alteration within a river network. Regul. Rivers Res. Manage. 14, 329–340. Richter, B.D., Baumgartner, J.F., Powell, J., Braun, D.P., 1996. A method for assessing hydrologic alterations within ecosystems. Conserv. Biol. 10, 1163–1174. Richter, B.D., Baumgartner, J.F., Wigington, R., Braun, D.P., 1997. How much water does a river need? Freshw. Biol. 37, 231–249. Richter, B.D., Richter, H.E., 2000. Prescribing ﬂood regimes to sustain riparian ecosystems along meandering rivers. Conserv. Biol. 14, 1467–1478. Robinson, C.T., Tockner, K., Ward, J.V., 2002. The fauna of dynamic riverine landscapes. Freshw. Biol. 47, 661–677. Rood, S.B., Mahoney, J.M., 2000. Revised instream ﬂow regulation enables cottonwood recruitment along the St. Mary River, Alberta, Canada. Rivers 7, 109–125. Rood, S.B., Samuelson, G.M., Braatne, J.H., et al., 2005. Managing river ﬂows to restore ﬂoodplain forests. Front Ecol. Environ. 3, 193–201. Rubin, D.M., Topping, D.J., 2001. Quantifying the relative importance of ﬂow regulation and grain size regulation of suspended sediment transport ‘‘and tracking changes in grain size of bed sediment’’. Water Resour. Res. 37, 133–146. Schmedtje, U., 1995. Beziehungen zwischen der sohlnahen Stro¨mung, dem Gewa¨sserbett und dem Makrozoobenthos in FlieXgewa¨ssern – O¨kologische Grundlagen fu¨r die Beurteilung von Ausleitungsstrecken. Institut fu¨r Zoologie und Limnologie, Universita¨t Innsbruck, Innsbruck, Austria. Schneider, M., Jorde, K., 2003. Fuzzy-rule based models for the evaluation of ﬁsh habitat quality and instream ﬂow assessment. In: Lamb, B.L., Garcia de Jalon, D., Sabaton, C., Souchon, Y., Tamai, N., Robinette, H.R., Waddle, T.J., and Brinson, A. (Eds), International IFIM Users’ Workshop. Fort Collins,, CO, USA. Snyder, E. and Minshall, G., 1996. Ecosystem metabolism and nutrient dynamics in the Kootenai River in relation to impoundment and ﬂow enhancement for ﬁsheries management. Final Report. Stream Ecology Center, Idaho State University, Pocatello, ID. Statzner, B., Gore, J.A., Resh, V.H., 1988. Hydraulic stream ecology: observed patterns and potential applications. J. N. Am. Benthol. Soc. 7, 307–360. Tetratech, 2004. Kootenai River geomorphic assessment – Final Report. U.S. Army Corps of Engineers, Seattle District, Seattle, WA. Tockner, K., Ward, J.V., Edwards, P.J., Kollmann, J., 2002. Riverine landscapes: an introduction. Freshw. Biol. 47, 497–500.

636

K. Jorde et al.

Townsend, C.R., 1989. The patch-dynamics concept of stream community ecology. J. N. Am. Benthol. Soc. 8, 36–54. USACE, 1984. Libby Dam Water Control Manual. U.S. Army Corps of Engineers, Seattle District, Seattle, WA. Vannote, R.L., Minshall, G.W., Cummins, K.W., et al., 1980. The River Continuum Concept. Can. J. Fish. Aquat. Sci. 37, 130–137. Vekerdy, Z., Meijerink, A.M.J., 1998. Statistical and analytical study of the propagation of ﬂood-induced groundwater rise in an alluvial aquifer. J. Hydrol. 205, 112–125. Ward, J.V., 1989. The 4-dimensional nature of lotic ecosystems. J. N. Am. Benthol. Soc. 8, 2–8. Ward, J.V., Stanford, J.A., 1983a. The intermediate-disturbance hypothesis: an explanation for biotic diversity patterns in lotic ecosystems. In: Fontaine, T.D. and Bartell, S.M. (Eds), Dynamics of Lotic Ecosytems, Ann Arbor Science, Ann Arbor. Ward, J. V., Stanford, J.A., 1983b. The serial discontinuity concept of lotic ecosystems. In: Fontaine, T.D. and Bartell, S.M. (Eds), Dynamics of Lotic Ecosystems, Ann Arbor Science, pp. 29–42. Ward, J.V., Stanford, J.A., 1995. The serial discontinuity concept: extending the model to ﬂoodplain rivers. Regul. Rivers Res. Manage. 10, 159–168. Ward, J.V., Tockner, K., Arscott, D.B., Claret, C., 2002. Riverine landscape diversity. Freshw. Biol. 47, 517–539. Wiens, J.A., 2002. Riverine landscapes: taking landscape ecology into the water. Freshw. Biol. 47, 501–515. Williams, G.P. and Wolman, M.G., 1984. Downstream effects of dams on alluvial rivers. U.S. Geological Survey Professional Paper, Washington, DC. Zelch, K., 2003. Aggrading Alluvial Fans and their Impact on Fish Passage in Tributaries of the Kootenai River, Idaho and Montana. College of Natural Resources. University of Idaho, Moscow, ID.

Gravel-Bed Rivers VI: From Process Understanding to River Restoration H. Habersack, H. Pie´gay, M. Rinaldi, Editors r 2008 Elsevier B.V. All rights reserved.

637

24 Movements of a macroinvertebrate (Potamophylax latipennis) across a gravel-bed substrate: effects of local hydraulics and micro-topography under increasing discharge Stephen P. Rice, Thomas Bufﬁn-Be´langer, Jill Lancaster and Ian Reid

Abstract Flow refugia provide a mechanism that can explain the persistence of macroinvertebrate communities in ﬂood-prone, gravel-bed rivers. The movement behaviour of macroinvertebrates is a key element of the ﬂow refugia hypothesis, but surprisingly little is known about it. In particular, little is known about how local near-bed hydraulics and bed micro-topography affect macroinvertebrate movements. We used a novel casting technique to reproduce a natural gravel-bed substrate in a large ﬂume where we were able to observe the movement behaviour of the cased caddisﬂy, Potamophylax latipennis at different discharges. The crawling paths and drift events of animals were analysed from video recordings and used to classify sites on the substrate according to the type of insect movement. We used acoustic Doppler velocimeter (ADV) measurements close to the boundary to characterise hydraulic conditions at different sites and a detailed Digital Elevation Model (DEM) to characterise sites topographically. Animals made shorter more disjointed crawling journeys as discharge increased, although they tended to follow consistent paths across the substrate. As hypothesised, crawling behaviour was locally associated with low elevations, low ﬂow velocities and low turbulent kinetic energies, while sites that insects avoided were characterised by higher elevations, velocities and turbulence. Discrimination was greater at higher discharges, indicating that movement behaviour is contingent upon ﬂow conditions. We suppose that these relations reﬂect the need of animals to reduce the risk of entrainment and minimise energy expenditure by avoiding areas of high ﬂuid drag. As discharge increased, there was a general upward shift in the frequency distributions of local velocities and turbulent kinetic energies. The animals responded to these shifts and it is clear that their different activities were not limited to ﬁxed ranges of velocity and turbulence. We assume that the absolute hydraulic forces would become a limiting factor at some higher discharge. At the discharges examined here, which are below those required to instigate framework E-mail address: [email protected] (S.P. Rice) ISSN: 0928-2025

DOI: 10.1016/S0928-2025(07)11152-4

S.P. Rice et al.

638

particle entrainment, patterns of animal movement appear to be associated with the animals’ experiences of relative rather than absolute hydraulic forces. 1.

Introduction

How do populations of stream invertebrates persist in gravel-bed rivers subject to hydrological disturbances (ﬂoods) where hydraulic forces and associated bed material movement can result in mortality and/or loss of individuals? This is a problem of enduring interest to stream ecologists and an area for fruitful collaboration between ecologists and geomorphologists. Understanding this problem involves three linked ideas. (1) In gravel-bed rivers, near-bed hydraulics are conditioned by the complex threedimensional micro-topography of the bed materials and are therefore spatially heterogeneous. This heterogeneity persists at the highest ﬂows and some lowstress areas are present even at high discharges. Spatially distributed ﬂow measurements above gravel beds are rare (Lamarre and Roy, 2005) and the degree to which spatial heterogeneity is maintained as discharge varies has not been widely studied or quantiﬁed. Nevertheless, the persistence of low-stress areas across a range of discharges has been demonstrated, both at ﬁxed locations (Lancaster and Hildrew, 1993a) and at shifting locations in association with changing stage (Rempel et al., 1999). (2) The distribution of benthic invertebrates across the stream bed is also spatially and temporally heterogeneous and this patchiness is associated with the heterogeneity in near-bed hydraulics. Several studies have shown that during high ﬂows, animal densities are higher in areas of low shear stress and low velocity (Lancaster and Hildrew, 1993b; Palmer et al., 1996; Rempel et al., 1999). Additional abiotic and biotic factors (substrate, food availability, predation, competition) contribute to patchy invertebrate organisation, but a wealth of evidence suggests that ﬂow is a primary consideration (e.g., Hart and Finelli, 1999) by direct inﬂuence on entrainment and by indirect effects, such as the availability of particulate organic foodstuffs (Bouekaert and Davis, 1998). (3) During ﬂoods, parts of the stream bed that experience low hydraulic stresses may act as ﬂow refugia, such that invertebrates that happen to be in, or move into, these areas avoid entrainment. Local population loss is thereby minimised and a group of survivors is available to re-colonise the bed. Flow refugia have been associated with single stable stones (Matthaei et al., 2000), microform bed clusters (Biggs et al., 1997), the hyporheos (Dole-Olivier et al., 1997), bar edges (Rempel et al., 1999), inundated ﬂoodplains (Badri et al., 1987) and woody debris (Palmer et al., 1996), as well as undifferentiated in-channel zones of relatively low velocity (Lancaster and Hildrew, 1993a). The relative importance of these refugia is contested (Palmer et al., 1992; Robertson et al., 1997; Matthaei and Townsend, 2000; Matthaei and Huber, 2002), and it seems most likely that they serve different animals at different times, depending on their availability and the life stage and traits of the animals.

Movements of a macroinvertebrate across a gravel-bed substrate

639

In all cases, however, the efﬁciency of the ﬂow refugia mechanism relies upon the passive or active movement of animals into and, perhaps, out of the protected areas. Thus, important keys to understanding the basis of how ﬂow refugia can facilitate population persistence are: an understanding of the movement behaviour of invertebrates across the stream bed; how this behaviour is inﬂuenced by local hydraulic conditions; and the net effect of those movements at the population level (Lancaster and Belyea, 1997). Surprisingly little is known about the movement of benthic macroinvertebrates in natural settings; for example, there is almost no information about the velocity at which insect larvae are able to move across a gravel substrate, their preferred pathways of movement in relation to bed micro-topography and whether movement behaviour changes in response to changes in the general ﬂow characteristics. The effects of stream ﬂow on invertebrate drift have received some attention (e.g., recent review in Hart and Finelli, 1999), but detailed studies of movements in association with the substrate (crawling, walking) are scarce. There is indirect empirical evidence of the role hydraulics plays in the dynamic spatial distribution of invertebrates from ﬁeld surveys (Lancaster and Hildrew, 1993b; Palmer et al., 1996; Rempel et al., 1999; Lancaster and Belyea, 2006) and some manipulative ﬁeld experiments (Winterbottom et al., 1997; Lancaster, 2000), but there is a lack of direct observations of invertebrate movement in realistic environments. This is the general focus of our work. In this paper, the interaction between insect movement, local micro-topography and hydraulics at the stream bed are investigated. We map the paths taken by insects as they move across a realistic facsimile of a gravel surface above which near-bed ﬂow is spatially heterogeneous. We then examine the differences in hydraulics and elevation at locations where different types of movement and different levels of mobility are observed. In another paper, we examine how these interactions might have higher order implications for population-level processes, e.g., net displacement, spatial dispersion, etc. (Lancaster et al., 2006).

2.

Rationale, aims and approach

Our basic assumptions are that animals seek to minimise the energy costs of movement (Vogel, 1981; Huryn and Denny, 1997) and minimise the risk of long-distance, downstream displacement through entrainment and drift. Drift is often regarded as a surrogate measure of mortality because it may result in increased predation risk, reduced feeding opportunities, physical damage and/or transport to unsuitable habitat (Palmer et al., 1992, 1996). A few individuals may survive long distance drift in the order of thousands of metres, including some vagile mayﬂies (Hershey et al., 1993) and cased caddis that rarely drift (Neves, 1979), but catastrophic or unintentional drift is still likely to be a high risk activity. We therefore expect moving animals to avoid locations that are energetically expensive or where the risk of entrainment is high in favour of sites where energy expenditure and the risk of entrainment are low. Sites that experience high ﬂow velocity are likely to entail greater energy expenditure because animals have to work harder to resist and overcome drag. Sites that experience high turbulent kinetic energy probably exhibit a

640

S.P. Rice et al.

higher risk of entrainment because animals are subject to more variable and more extreme ﬂuid forces, increasing the probability that entrainment thresholds are exceeded. It is therefore likely that high-risk sites are characterised by high near-bed velocity and high turbulence intensity, whereas low risk sites exhibit low near-bed velocity and low turbulence intensity. Notwithstanding the complex three-dimensional ﬂow ﬁelds created by a rough gravel boundary, greater elevation implies greater exposure to higher velocities and exclusion from dead-water zones, for example at interstitial junctions. So, differences in macroinvertebrate behaviour with local elevation are also likely. Speciﬁcally, we hypothesise that: Hypothesis 1. Sites characterised according to type of movement activity will have signiﬁcantly different near-bed hydraulics, such that velocity and turbulence intensity are lower at sites where crawling is common, but higher at sites associated with entrainment. Places where there is little or no insect activity, to the extent that they can be regarded as sites that are ‘avoided’ or less favourable, are expected to exhibit the highest velocities and turbulent kinetic energies. Hypothesis 2. Sites characterised according to type of movement activity will have signiﬁcantly different local elevations such that crawling is common at relatively low average elevations and entrainment occurs at higher elevations. We assume that any effects due to elevation reﬂect local hydraulic differences, and view assessment of Hypothesis 2 as an extension of Hypothesis 1, acknowledging that the more easily obtained elevation data are used as a simple surrogate for more hard-won hydraulic information. In another paper, we have shown how the hydraulic environment close to a natural gravel surface changes with the general ﬂow condition. In particular, we have demonstrated that the increases in spatially averaged velocity and turbulence intensity that are driven by changes in channel discharge are associated with increased local values and greater spatial variability of these ﬂow parameters (Bufﬁn-Be´langer et al., 2006). Here, we investigate whether such changes affect the relations between macroinvertebrate movement, local hydraulics and local topography by assessing Hypotheses 1 and 2 across three discharge conditions. Higher discharges may be associated with greater hydraulic differentiation of sites where different activities occur, because conditions are more constraining and local environments are more diverse, and we hypothesise that: Hypothesis 3. Differences in the hydraulics and/or elevation of sites where contrasting movement activities occur will become greater as discharge increases. We therefore consider the role of both highly local hydraulics, as controlled by the micro-topography of the gravel surface, and the effects of general increases in ﬂow. Detailed measurements of near-bed hydraulics and macroinvertebrate movements on gravel substrates are difﬁcult to achieve in the ﬁeld, becoming impracticable at high ﬂows. Conversely, controlled measurements are possible in laboratory ﬂumes, but the reproduction of natural gravel fabrics requires greater transport rates than

Movements of a macroinvertebrate across a gravel-bed substrate

641

can be generated artiﬁcially, limiting the realism and validity of the data obtained. Previous hydraulic studies have attempted to reconstruct (Young, 1992), import (Bufﬁn-Be´langer, 2001) or reproduce (Lawless and Robert, 2001) natural gravel structures in the laboratory, or have developed water-worked textures using small gravels (Kirkbride, 1993; Lane et al., 2004). Small scale (cm) movements of individual invertebrates are extremely difﬁcult to observe directly in the ﬁeld. Those studies which have done so have necessarily avoided natural, high ﬂows and worked at very small scales on individual clasts (Poff and Ward, 1992; Hart et al., 1996) or have worked at larger scales with less detailed behavioural observations and little hydraulic detail (Hart and Resh, 1980; Jackson et al., 1999). Most previous detailed studies of insect movement-ﬂow interactions have therefore used ﬂumes with highly simpliﬁed environments or a random arrangement of gravels (e.g., Holomuzki and Biggs, 1999; Lancaster, 1999). Accordingly, the nearbed hydraulics of these experimental arenas are likely to be unrepresentative of natural stream channels. To overcome these problems, a novel casting technique (Bufﬁn-Be´langer et al., 2003) was used to produce a precise replica of a ﬂuvial cobblegravel substrate that was deployed in a large laboratory ﬂume, where we could control and manipulate ﬂow, take detailed hydraulic measurements and observe invertebrate movements.

3. 3.1.

Methods Experimental arrangement

A precise replica of a natural cobble-gravel substrate, 1.0 by 2.0 m, was made using the casting technique described by Bufﬁn-Be´langer et al. (2003). The cast was obtained from an exposed gravel bar in the River Manifold, UK, and reproduces the true three-dimensional complexity of a natural, water-worked unit. The cast retains signiﬁcant small-scale details including the texture of mosses and sands and most of the surface interstices. Orthophotographs and a digital elevation model (DEM) of the cast surface were generated by close-range digital photogrammetry (Chandler et al., 2003). A grid-by-number grain size distribution obtained by a non-invasive photographic technique (Graham et al., 2005) and truncated at 0.008 m, yields a median diameter D50 ¼ 0.048 m and a D95 ¼ 0.119 m. A representative sub-area of the cast 1.1 m long and 0.80 m wide was selected for detailed hydraulic measurements and insect observations (Fig. 24.1). This area is many times the area of the individual grains that make up the bed. Elevation data for this area were examined to characterise surface roughness. Elevations, h (n ¼ 38,480), measured relative to the lowest point in the sub-area, rise to 0.121 m, have a median h50 ¼ 0.053 m, are positively skewed and exhibit a lognormal distribution. Comparison of this elevation distribution with previously published data indicates that the cast and the prototype gravel patch are representative of water-lain gravel surfaces in other rivers. For example, the lognormal ﬁt is consistent with the ﬁeld observations of Smart et al. (2004) who made detailed roughness characterizations of six natural river gravels. Also, the

642

S.P. Rice et al.

Figure 24.1. Topography of the cast surface (0.8 1.1 m). The x-axis referred to in the text is the long, streamwise axis, y is cross-stream and z is vertical. All axis units are cm. (a) Oblique view of the DEM looking from the right bank, ﬂow left to right. Three prominent features are indicated: the large cobble located toward the downstream end (A); the low-lying plane-bed area to its left (B) and the upstream particle of an imbricate cluster, centre-right (C). (b) Contour map of the DEM showing the ﬁve seeding locations (MH6–MH10) and the 110 ADV sampling positions (solid black dots). (c) Oblique photograph looking upstream from the left bank (note pen-top centre at C-for scale).

skewness value hSK ¼ 0.46 is similar to the average value reported by Nikora et al. (1998) for 77 ﬁeld proﬁles from eight gravel-bed rivers (hSK ¼ 0.47, SD ¼ 0.51) and notably different from the negative skewness values reported for manually created, ‘unworked’ ﬂume beds (Kirchner et al., 1990). The cast was positioned in a 9.0 m long, 0.9 m wide and 0.8 m deep ﬂume with a ﬁxed slope of 0.002. Full details of the ﬂume set-up can be found in Bufﬁn-Be´langer et al. (2006). Three uniform and steady ﬂows were established by running the ﬂume successively at three discharges while minimising changes in water depth (Table 24.1). All three ﬂows were fully turbulent, sub-critical, representative of ﬂows in natural rivers and differentiated by increases in a variety of relevant ﬂow characteristics (e.g., mean velocity, Reynolds number, shear velocity). The three ﬂows represent a distinct treatment (referred to as Flow in our analyses) that provides a means of examining

Movements of a macroinvertebrate across a gravel-bed substrate Table 24.1.

643

Characteristics of the three experimental ﬂows.

Flow

Q (m3 s1)

C0.4 (m s1)

Y50 (m)

Re

Fr

tbed (N m2)

v* (m s1)

1 2 3

0.153 0.202 0.262

0.310 0.498 0.764

0.536 0.439 0.368

166160 218622 281152

0.135 0.240 0.402

0.530 1.467 3.693

0.023 0.038 0.061

Q is discharge, C0.4 an average representative velocity at 0.4Y50, Y50 the median water depth, i.e., the water depth above h50, Re is Reynolds number, Fr is Froude number, tbed the reference bed shear stress and v* the shear velocity. tbed and v* are estimated from velocity proﬁles taken upstream from the cast using the law of the wall [v ¼ 2.5v*(ln(y/y0)) and tbed ¼ r(v*)2] applied to velocity measurements in the near-bed region (below 0.4Y50).

the interactive effects of local hydraulics and general, externally driven, ﬂow conditions that, in some respects, mimic the rising limb of a ﬂood hydrograph. Gravel-bed rivers are characterised by hydraulic stresses that rarely exceed the entrainment thresholds of the framework particles exposed at the bed surface, so that insects are often exposed to high ﬂows but beneﬁt from a stable substrate. Estimated bed shear stress for the highest discharge was below the critical entrainment threshold of the framework particles in the prototype gravel patch, so the ﬂows used are consistent with the stability of this particular substrate. The ﬁxed nature of the substrate surface was important for our experiments because it meant that we were able to examine the effects of ﬂow forces only on insect movement, without the confounding effects of substrate movement.

3.2.

Hydraulic measurements

For each ﬂow, spatially-distributed, near-bed hydraulic measurements were made using an acoustic Doppler velocimeter (ADV). The ADV was deployed at 110 locations in an 11 by 10 x–y grid with spacings of 0.1 and 0.05 m, respectively (Fig. 24.1b). At each location, velocity measurements were made at z ¼ 0.008 m above the local bed, rather than above an arbitrary horizontal reference plane. The full set of spatially distributed measurements therefore describes the hydraulics in a convolute layer that follows the topographic highs and lows of the surface. The sampling volume of the ADV is cylindrical and less than 200 mm3. The near-bed positions, the sampling volume and the use of local topography as the reference height ensured that our measurements describe the hydraulic conditions experienced by macroinvertebrates moving across the gravel surface. At each location, instantaneous velocities were measured for the three orthogonal velocity components (streamwise, U; cross-stream, W; vertical, V) over a period of 60 s at a sampling frequency of 25 Hz. The potential sources of error in ADV data are well-understood (Lane et al., 1998; Nikora and Goring, 1998; Finelli et al., 1999; McLelland and Nicholas, 2000; Wahl, 2000) and a rigorous validation scheme was employed to ensure maximum quality (Bufﬁn-Be´langer et al., 2006). The mean and standard deviation for each velocity component (e.g., /US, URMS) were extracted

S.P. Rice et al.

644

from the velocity time series at each location and turbulent kinetic energy, a surrogate for the intensity of three-dimensional turbulent ﬂuctuations about the mean ﬂow, was computed as: K ¼ 0:5rðU 2RMS þ V 2RMS þ W 2RMS Þ

(24.1)

where r is water density ( ¼ 1000 kg m3) and K is in J m3.

3.3.

Macroinvertebrate observations

Larvae of the cased caddisﬂy, Potamophylax latipennis (Curtis) were selected for use in the experiments. At instars IV and V, these insects are relatively large, slowmoving, benthic species that live inside a cylindrical case built from medium-coarse sand grains. The average case had a length of 20 mm (range 17–28 mm) and an average density of approximately 1100 kg m3. Further methodological details regarding the invertebrate observations can be found in Lancaster et al. (2006). In each of six replicate trials, carried out at the three different and predetermined ﬂow conditions, we recorded the movement of ﬁve animals. Five points located within the ADV measurement grid were selected as seeding or start locations (MH6–10; Fig. 24.1b). Seeding locations were relatively sheltered positions, usually in the lee of particles that projected slightly above the general surface, where the larvae could be introduced with minimal danger of immediate entrainment. Using ﬁve animals per trial minimised the net duration of the experiment and use of ﬁxed seeding locations maximised the likelihood that animals would settle on the cast. The periods of observation for a trial ranged between 1 and 23 min with an average of approximately 9 min. The experiment was a split-plot design with two fully orthogonal ﬁxed factors (ﬂow, seeding location) and one random factor (trial) nested within ﬂow treatment. Larval movements were recorded using a digital video camera suspended above the substrate surface. A Perspex viewing box was positioned carefully on the water surface without altering signiﬁcantly the near bed hydraulics in order to facilitate an undistorted image of the substrate. The video imagery has dimensions of 767 by 575 pixels and covers an area of approximately 1 m2, giving a ground resolution of 0.0015 m. From the video recordings we obtained the x and y location of each caddis every 5 s and also the time and x and y locations of entrainment start and stop (reattachment) positions. These time series of (x, y) coordinates deﬁne a trace across the cast surface of the path followed by each larva. At each interrogation (5 s intervals), the behaviour of the larva was classiﬁed as crawling, entrained, struggling or stationary. Larvae never truly drifted, i.e., were never fully suspended in the water column, but rather they tumbled close to the bed surface. Relevant metrics estimated for each larva included proportion of time crawling, average crawling velocity and total displacement by entrainment. A full analysis of these metrics is reported elsewhere (Lancaster et al., 2006), but general results are used here to support our investigation.

Movements of a macroinvertebrate across a gravel-bed substrate 3.4.

645

Data analysis

A set of screws was embedded in the prototype bar, reproduced in the cast and used as spatially distributed benchmarks. It was then possible to locate DEM postings, orthophotographs, larval paths and hydraulic data within a single coordinate system that allowed extraction of spatially explicit information. A radial correction was applied to the larval paths to allow for distortion in the video imagery due to the camera lens. Remaining errors in the video trace data due to camera tilt and camera rotation are small and we estimate a maximum cumulative error in the positioning of insects of 0.0075 m. Crawling and drift paths for all larvae were grouped by Flow and mapped onto orthophotographs and three-dimensional renderings of the DEM. These maps provided a means of examining general patterns and, in addition, allowed locations on the cast surface to be classiﬁed with respect to four categories of insect activity, which we subsequently refer to by the factor name Activity: crawling (one or two crawling paths), congested crawling (more than two crawling paths), entrainment sites and sites of no activity (animals were never present, so no crawling or entrainment were recorded). This classiﬁcation scheme was applied to circular areas of the bed (radius ¼ 20 mm), which we call ‘‘sites’’, centred on each of the ADV sampling positions. Some sites were excluded from analyses a priori, e.g., sites where estimates of hydraulic conditions were impossible and seeding locations where high insect activity simply reﬂected the sampling design. Differences in hydraulic characteristics at 0.008 m above the local bed elevation (/US, K ) between insect activities (Activity; Hypothesis 1) and across the ﬂow treatments (Flow; Hypothesis 3) were then examined using two-way ANOVA. The elevations at which these different activities took place were examined in a similar manner. Crawling and entrainment elevations were obtained by extracting values of h from the DEM along the crawl path and at points of entrainment. Crawling elevations during each ﬂow were compared to the mean elevation of the sampling sub-area of the cast surface using one-sided, one-sample t-tests. Entrainment and crawling elevations were compared at each ﬂow using paired t-tests with the mean entrainment elevation of each individual compared with its own mean crawling elevation (Activity; Hypothesis 2). We tested for the effect of Flow (Hypothesis 3) on crawling elevation using a split-plot ANOVA (ﬁxed factors ¼ Flow, seeding location; random factor ¼ trial). The design assumes that there is no signiﬁcant interaction term, Location Trial (Flow), and the residual error cannot be estimated (MS for the random factor is used as the error term in the F-test). In no case were any interaction terms signiﬁcant so, for brevity in the text that follows, only the main Flow effects are reported.

4. 4.1.

Results Path Maps

The path maps provide a clear illustration of how insect movements were affected by ﬂow. As discharge increased (Fig. 24.2, panels a–c), crawling paths became shorter

646

S.P. Rice et al.

and increasingly disjointed. The average straight-line displacement resulting from individual crawling events declined signiﬁcantly, from 67 to 30 mm and then to 27 mm under ﬂows 1, 2 and 3, respectively (F2,15 ¼ 18.1, po0.001, Flow effect in split-plot ANOVA). In contrast, the frequency of entrainment events increased 23-fold from ﬂow 1 to ﬂow 3 and distance per event increased similarly. Thus, as discharge increased, long crawling journeys were replaced with shorter journeys and were broken by periods of inactivity or by entrainment (Lancaster et al., 2006). The path maps are also useful in revealing that there was a tendency for crawling individuals to converge on particular locations and to travel along the same routes as one another. Spatial concentration of crawling activity was inevitable close to the seeding sites, but beyond the immediate vicinity of the ﬁve seeding locations, certain pathways were frequently used, suggesting a degree of preference. The consistent use of particular corridors was apparent under each ﬂow condition (Fig. 24.2a–c) and between ﬂow conditions (Fig. 24.2d). For example, under ﬂows 1 and 2, a corridor of activity extended downstream from the seeding locations toward the right edge of the

Figure 24.2. Crawl traces (white lines) and the starting locations of entrainment events (black dots) for ﬂows 1, 2 and 3 (respectively a, b, c). A composite of the traces for the three ﬂows is shown in d (thick black is ﬂow 1, dark grey is ﬂow 2 and thin black is ﬂow 3). In each case, all replicates from each seeding location (black circles with a cross) are shown. The background image is an orthophotograph that is spatially consistent with the trace maps. Flow is from left to right and the large cobble (A in Fig. 24.1, with moss showing) is located toward the bottom right of each image.

Movements of a macroinvertebrate across a gravel-bed substrate

647

large cobble centred on x ¼ 60 cm, y ¼ 10 cm (labelled A in Fig. 24.1). At ﬂow 3, activity within this corridor was restricted to two small areas. Topographic and inferred hydraulic conditions provide possible reasons why the larvae converged on particular paths. For example, Figs. 24.2 and 24.3 show that a set of partially imbricated gravel particles form a weak cluster that ﬂanks the upstream edge of the corridor highlighted above, perhaps providing a hydraulically sheltered path in its lee where larvae could move with relative ease. In contrast, the large cobble just downstream has an exposed stoss face where no traces were recorded. The consistent movement of animals toward its right-hand edge at all discharges might indicate that the stoss face was unacceptable crawling terrain or simply that the cluster-protected corridor presented an acceptable route that did not require the animals to seek alternative paths. At ﬂows 2 and 3, movements were also noted along the upstream edge of the cobble, toward its left ﬂank and, indeed, at ﬂow 3 the outline of the cobble’s base was essentially traced out by crawling activity of multiple individuals (Fig. 24.2c). Whether by design or by default, the larvae tended not to crawl on top of this or other large exposed particles, particularly at stronger ﬂows (Fig. 24.3). Thus, at ﬂows 1 and 2 several paths crossed the upper surface of the particle centred at x ¼ 25 cm, y ¼ 0 cm but, at ﬂow 3, no paths are apparent on this surface. The path maps (Figs. 24.2 and 24.3) suggest that larvae tended to crawl around large particles rather than over higher, exposed surfaces. This is consistent with our observations which show that, when crawling, larvae were most frequently observed on plane-bed areas (where the upper surfaces of adjacent grains lie at approximately the same elevation) or at the interstitial junctions between plane beds and larger, taller clasts (75–85% of the time under ﬂow 1; 61% under ﬂow 3) rather than on particle tops or sides (Lancaster et al., 2006).

4.2.

Site hydraulics, general flow conditions and insect activity

Site hydraulics inﬂuenced insect activity and there was an interactive effect in that the strength of this inﬂuence varied with the general ﬂow condition (ﬂow 1, 2 or 3), i.e., insect movement behaviour changed with ﬂow. Fig. 24.4 shows the mean and the 95% conﬁdence limits of /US and K for sites characterised by congested crawling, crawling, entrainment and no activity, under the three ﬂow treatments. At the majority of sites there was no activity (60–70%), but the frequency of no activity sites did not vary with ﬂow (w2 ¼ 1:24, p40:05). For both of the hydraulic variables, velocity and turbulence, the main treatments of Flow and Activity were signiﬁcant and the interaction term was signiﬁcant for velocity (Table 24.2). An increase in site velocity and turbulence with increasing ﬂow is unsurprising. However, noteworthy in terms of Activity, is that velocity and turbulence were lower in congested crawling sites and higher in no activity sites (Fig. 24.4). The hydraulic character of crawling and entrainment sites generally lay between the extreme conditions, i.e., those typical of no activity and congested crawling sites. The signiﬁcant interaction between Flow and Activity for velocity is particularly interesting (i.e., movement activity is contingent upon ﬂow) and attributed to an increase in the magnitude of the difference in

648

S.P. Rice et al.

Figure 24.3. Oblique view of the DEM from the right bank with crawl traces under (a) ﬂow 1 and (b) ﬂow 3 showing the topography associated with favoured pathways. Flow is from left to right and the large cobble (A in Fig. 24.1) is located toward the bottom right of each image.

site hydraulics for each type of activity as discharge increases (Table 24.2). Thus, sites of different activity are most clearly differentiated by velocity under ﬂow 3. Indeed, it is apparent that the signiﬁcant Activity effect in the ANOVA results is largely due to the differences in velocity between the activity groups at Flow 3 (Table 24.2). While the general trend from low velocity at congested crawling sites to higher velocity at no activity sites is consistent at the two other discharges (Fig. 24.4), the differences are not statistically signiﬁcant (Table 24.2). This suggests that animals strongly avoided the highest velocity sites at high ﬂow, but were less discriminating at lower ﬂow (Fig. 24.4).

Movements of a macroinvertebrate across a gravel-bed substrate 0.5

649

12

0.4

Turbulent Kinetic Energy, K (J m−3)

Mean Velocity, (m s−1)

crawling congested crawling entrainment no activity

0.3 0.2 0.1 0

10

-0.1

8 6 4 2 0

1

2 Flow

3

1

2 Flow

3

Figure 24.4. Mean and 95% conﬁdence limits of time-averaged streamwise velocity, /US, and turbulent kinetic energy, K, for crawling, congested crawling, entrainment and no activity sites under each ﬂow condition. Hydraulic measurements were obtained at a height of 0.008 m above the local bed surface. See Table 24.2 for summary of statistical analysis.

Table 24.2. Summary of two-way ANOVA testing for differences in hydraulic conditions (mean streamwise velocity and turbulent kinetic energy) at sites classiﬁed according to insect activity (Activity) and discharge (Flow). Variable

Source

df

MS

F

P

(a) Velocity /US

Flow Activity Flow 1 Activity Flow 2 Activity Flow 3 Activity Flow Activity Residual

2 3 3 3 3 6 277

0.0213 0.0139 0.0009 0.0016 0.0192 0.0051 0.0018

11.9 7.78 0.520 0.865 10.7 2.86

o0.001 o0.001 0.669 0.460 o0.001 0.010

(b) TKE, K

Flow Activity Flow Activity Residual

2 3 6 277

5.38 0.142 0.013 0.030

182 4.81 0.428

o0.001 0.003 0.860

Given the signiﬁcant interaction term in (a), we tested the effect of activity separately for each ﬂow, with MSresidual of the full model as the denominator. Transformation of log(x+1) for velocity and log(x) for K was carried out to meet assumptions of normality. Unbalanced replication is accounted for (no association between within-cell variance and sample size). To adjust for the missing cell (entrainment was rare at ﬂow 1), we nominally included one site in which there was entrainment, but that had been excluded a priori (see text for explanation). Thus the need to avoid a missing cell in the analysis out-weighed the cautious data screening. See Fig. 24.5 for illustration.

S.P. Rice et al.

650

These results suggest that crawling is associated with lower local velocities and lower turbulence, but that this activity is not restricted to a narrow range of hydraulic conditions. In particular, the results suggest that larvae are able to respond to shifts in the range of local hydraulic characteristics driven by general changes in discharge. 4.3.

Elevation, general flow conditions and insect activity

The path maps and the simple metrics, reported above, imply that larvae tended to crawl at relatively low elevations and this was further supported by a signiﬁcant difference between the average elevation of the cast surface and the mean crawling elevation at ﬂow 3 (t26 ¼ 5.39, po0.001) (Fig. 24.5). For ﬂows 1 and 2, there was a strong suggestion that crawling elevation was less than the mean surface elevation (t30 ¼ 1.62, p ¼ 0.059 and t25 ¼ 1.64, p ¼ 0.057, respectively). As discharge increased, however, there was no change in the mean crawling elevation, even though it appeared to be lower at ﬂow 3 (main ﬂow effect in split-plot ANOVA: F2,15 ¼ 1.60, p ¼ 0.23). Comparing the elevations at which entrainment occurred with crawling elevations yielded inconclusive results (Fig. 24.5). Mean entrainment elevation was generally higher than mean crawling elevation for individuals at ﬂow 2 (paired t-test: t24 ¼ 2.35, p ¼ 0.034), but not at ﬂow 3 (t21 ¼ 0.514, p ¼ 0.613). Similar analysis for ﬂow 1 was not possible given the scarcity of entrainment events. On the basis of these equivocal results it would be imprudent to conclude that entrainment is more or less likely from higher elevations, especially since the visual impression in Fig. 24.2 is that entrainment occurs from a wide range of topographic positions. This apparent 75 crawling

70

entrainment cast surface

Elevation, h (mm)

65 60 55 50 45 40 1

2

3

Flow Figure 24.5. Comparison of average crawling elevations, entrainment elevations and mean surface elevation for three ﬂow discharges. Mean values and 95% conﬁdence limits are shown. Only 1 drift event occurred under ﬂow 1 and it is not indicated here.

Movements of a macroinvertebrate across a gravel-bed substrate

651

indeterminacy might, in part, reﬂect the changing dimensions of coherent ﬂow structures that arise from changes in discharge. Points of shear layer detachment and reattachment and patterns of deﬂection of high-speed ﬂuid will alter as ﬂow increases and decreases. Elsewhere we have shown that, as discharge increases from ﬂow 2 to 3, there is a decrease in the average elevation of maximum turbulent kinetic energy associated with shear layers extending downstream from the crests of protruding particles (Bufﬁn-Be´langer et al., 2006). If entrainment probabilities are increased as turbulent kinetic energy increases, this lowering of shear layers might help to explain why entrainment elevations decline, on average, at ﬂow 3.

5.

Discussion

Crawling paths and entrainment maps (Figs. 24.2 and 24.3) highlight the patchy distribution of P. latipennis larvae as they crawl across a rough, gravelly substrate. Individual paths do not criss-cross the substrate at random but exhibit a degree of co-location which suggests that there are preferred crawling tracks. Analysis of the hydraulics and elevation at sites where different activities occur provides some insights into these insect preferences that, it is assumed, reﬂect a requirement to minimise energy expenditure and reduce the risk of entrainment. Sites where crawling and congested crawling are common are interpreted as those that most favour movement, and we have hypothesised that such sites would be characterised by relatively low elevations, low velocities and low turbulent kinetic energies. We have found strong evidence to support these hypotheses (Figs. 24.4 and 24.5; Table 24.2), especially at higher discharges. We have also hypothesised that entrainment would occur from sites with a higher average elevation and from sites with higher velocities and more intense turbulence than crawling sites. While velocities at entrainment sites tend to be higher than at congested crawling and crawling sites, turbulent kinetic energies are not, in general, any greater (Fig. 24.4). Our ﬁndings regarding the elevation of entrainment sites are equivocal and Fig. 24.2 suggests that entrainment occurs from a wide range of topographic positions. There is certainly no evidence to suggest that entrainment occurs exclusively at exposed sites or only where turbulence intensity and velocity are particularly high. Several reasons may explain why this hypothesis has not been validated. First, the total number of entrainment events was relatively small (107 events for 39 animals) and it is likely that better characterisation of entrainment sites requires a larger number of observations under stronger ﬂows. Second, there is the possibility that entrainment events are initiated by short-term velocity ﬂuctuations that are not revealed by ﬂow parameters such as the time-averaged streamwise velocity (/US) and the turbulent kinetic energy (K). The entrainment of an insect may occur following the passage of an intense velocity event, such as a sweeping motion. These events have been linked with the transport of bed sediments (e.g., Drake et al., 1988; Sumer et al., 2003) and could also be associated with the entrainment of benthic insects. Concurrent data on entrainment and velocity are needed to examine this possibility and would usefully be augmented by ﬂow visualisation to investigate the effect of turbulent ﬂow structures on entrainment.

652

S.P. Rice et al.

Areas without any activity are interpreted as places that animals actively avoided or places that they were simply not carried to. Given our interest in elucidating the role of local hydraulics and micro-topography on insect movement, the distinction between these is not important because both imply a signiﬁcant physical control on movement. Sites of no activity might also indicate places that insects did not have the opportunity to reach because their initial seeding location was too far away. However, we think this is relatively unimportant, because larvae did journey across the whole of the cast surface and no restrictions on observation time were enforced. So, areas of no activity are interpreted as areas that the insects avoided, though avoidance could have been active or passive. We have hypothesised that such areas would, in general, exhibit the highest local velocities and turbulent kinetic energies because these reﬂect the greatest energy costs and entrainment risk for mobile insects. The hypothesis is supported by the signiﬁcant Activity effects in our examination of hydraulic variables and the generally higher mean velocity and turbulent kinetic energy values at no activity sites compared with other sites (Table 24.2; Fig. 24.4). Overall these results demonstrate that the movement activity of P. latipennis is conditioned by micro-scale hydraulic patchiness (Hypotheses 1 and 2). Animals discriminated between sites, avoiding areas where entrainment risk and the energetic cost of ﬂuid drag were high and moving more frequently in low-lying areas or other places where velocity and turbulent kinetic energy values were relatively low. Our results illustrate the importance of hydraulic patchiness for benthic mobility and the retention of animals at particular locations. They therefore demonstrate that rough, heterogeneous substrates which create hydraulic patchiness, are important for the provision of in-stream refugia – at least for relatively slow-moving, crawling species such as this caddisﬂy. An interesting question is, then, whether gravel texture can be used to quantify the quality of in-stream refugia. It is clear that high quality in-stream refugia will be characterised by sufﬁcient low-velocity sites, which in turn implies heterogeneity in the size and arrangement of bed particles. Bed material sorting indices and structural characteristics may then provide a reasonable means of assessing refugia potential. This will be the subject of a forthcoming paper that compares near-bed hydraulic variability and insect activity between substrates that have different textural characteristics. It is evident from Fig. 24.4 that the distinction between no activity and other sites is most clear at ﬂow 3, where discharge is highest and this is reﬂected in the signiﬁcant interaction term for Flow and Activity in Table 24.2. The effect on local hydraulics of a general increase in ﬂow (an increase in discharge) and the fact that macroinvertebrates change their behaviour in response to changes in ﬂow are fundamental to understanding animal behaviours during ﬂood events. For the cast used here, Fig. 24.6 illustrates the effect of increased discharge on local mean velocity and turbulent kinetic energy: there was an increase in the average values in response to a general upward shift of the bulk of the frequency distribution and an increase in variability between sites. The upward shift of the values for the bulk of the sites means that, as discharge increased, animals were increasingly exposed to larger absolute velocities and turbulent kinetic energies across the whole of the cast surface. Although some sites continued to experience very low velocities and turbulent kinetic energies, these sites became increasingly rare. It is clear from Fig. 24.4 that the

Movements of a macroinvertebrate across a gravel-bed substrate

0.8

20

K

15 0.6 0.4

10

0.2 5 0

Turbulent Kinetic Energy, K (J m−3)

Mean Velocity, (m s−1)

1.0

653

0

-0.2 1

2 Flow

3

Figure 24.6. Changes in local time-averaged streamwise velocity, /US, and turbulent kinetic energy, K, with increasing discharge (ﬂows 1 to 3; see Table 24.1 for details) in a convolute layer 0.008 m above the bed for the 110 ADV sampling positions (see Fig. 24.1b). Note the increase in both the local mean values and the spatial variability as discharge increases.

Potamophylax larvae were able to respond to these shifts and that their activities were not limited to ﬁxed ranges of velocity and turbulence. Thus, at ﬂow 3, larvae crawled at sites with signiﬁcantly higher turbulent kinetic energy and velocity than at ﬂow 1 and were active in hydraulic conditions which, at lower ﬂows, characterised sites where no activity was observed. However, animals did discriminate among the available hydraulic conditions (e.g., avoiding sites with the highest velocities) and the strength of this discrimination increased with ﬂow (Hypothesis 3), i.e., movement behaviour is contingent upon ﬂow. This might suggest that insects key into spatial variations in relative hydraulic conditions, not absolute velocities, at least for the range of ﬂows studied here. Fig. 24.7 shows that the spatial pattern of relative velocities remained very consistent between different discharges so that sites with above or below average velocity at one ﬂow experienced above or below average velocity at other ﬂows. It is apparent in Fig. 24.2d that there was a degree of spatial consistency in movement patterns too, albeit that movements were more restricted under higher discharges. Together, these observations suggest that patterns of animal movement are associated with particular sites, despite local changes in hydraulics, because certain sites consistently represent the same opportunities in terms of energy savings and riskaversion. At discharges sufﬁcient to mobilise the bed, these relations are unlikely to persist and there must be an absolute limit to the hydraulic forces that animals can withstand without being entrained. But at modest ﬂows that do not exceed the critical threshold for framework particle entrainment, our results suggest that patterns of P. latipennis movement and avoidance might reﬂect relative opportunities rather than absolute hydraulic forces per se.

S.P. Rice et al.

654

Figure 24.7. Relative mean velocity measured 0.008 m above the bed surface (calculated as the local velocity divided by the mean of all local velocities for a given ﬂow) for (a) ﬂow 1 and (b) ﬂow 3. Relative velocities are consistent between ﬂows, despite signiﬁcant differences in absolute velocity and this is indicated in (c) which shows the modulus of the difference in relative mean velocity between sites for ﬂows 1 and 3.

6.

Conclusion

The substrate of a gravel-bed river is a dangerous place to live, subject to large hydraulic forces and prone to instability during large ﬂoods, yet benthic fauna are typically diverse and abundant. Flow refugia mechanisms help to explain the persistence of macroinvertebrate communities in gravel-bed streams. The movement of insects is central to the refugia idea, but little is known about the nature of insect

Movements of a macroinvertebrate across a gravel-bed substrate

655

movements on rough substrates and the effect of increases in ﬂow or differences in substrate texture on movement patterns, pathways and characteristics. Our observations of P. latipennis lead us to the general conclusion that crawling journeys tend to follow low-lying paths characterised by low velocity and, to a lesser extent, low turbulence intensity. In contrast, animals avoid sites characterised by the highest velocities and turbulence intensities, especially at higher ﬂows. Importantly, the strength of this behavioural response is contingent upon ﬂow. These observations give us an insight into how these insects utilise the substrate to avoid entrainment and minimise energy expenditure and illustrate the importance of hydraulic patchiness for movement behaviour. The degree to which animals avoid entrainment and conserve energy will depend upon the heterogeneity of the substrate in terms of particle size and arrangement. By extension, this suggests that high quality in-stream refugia should be characterised by heterogeneous substrates. Our ﬁndings add to the understanding of the basic principles that underlie the refugia mechanism and thence the question of how invertebrate communities persist in the face of hydraulic disturbances. In addition to rain and snow-melt ﬂooding, such disturbances are also associated with controlled water releases on managed gravel-bed rivers. In this respect, insights into the successful operation of refugia mechanisms should be a consideration of integrated river management because the survival of viable populations of macroinvertebrates is a vital element of the biotic well-being of any gravel-bed river. This paper has focused on linking near-bed hydraulics with the movement behaviour of a particular cased caddisﬂy at discharge levels below those that would mobilise the framework gravels. This approach, necessarily, did not consider three other sets of factors that are important for understanding macroinvertebrate movement behaviour in the context of refugia utilisation. First, a number of biotic factors are also likely to be important. Minimising energy expenditure and drift are not the only concerns of benthic insects. Biological controls might include movement strategies that are intended to maximise food acquisition, minimise the risk of predation or reduce competition between individuals. For example, crawling at low elevation along interstitial junctions may in part reﬂect avoidance of predation. Second, our results pertain to a single crawling species, but the morphological and behavioural traits of different macroinvertebrate species are of fundamental importance in terms of their movement characteristics so that it is difﬁcult to generalise our results to other species or to whole communities. For example, the shape of a species partly determines its Reynolds number and drag coefﬁcient and thence the ﬂuid forces that it is subjected to and its ability to resist entrainment (Vogel, 1981; Statzner and Holm, 1989) and move around. For larvae that live in cases there is the additional effect of the case’s weight. Waringer (1993) determined Reynolds numbers and drag coefﬁcients for a variety of dead, cased caddisﬂy and estimated critical entrainment stresses. Potamophylax cingulatus (Steph), which is similar to the species used here, had the highest Reynolds number and lowest drag coefﬁcient of the macroinvertebrates examined and required the highest tractive forces to entrain it. There is, therefore, a need to examine the movement of animals that possess contrasting characteristics in terms of size, weight and shape. This study indicates the experimental and analytical tools that might facilitate such work.

656

S.P. Rice et al.

Third, the role of bed stability in deﬁning suitable refugia is not considered in our experiments. Not surprisingly, given the difﬁculties of making reasonable observations, work on insect distributions and mortality across mobile beds is scarce (Holomuzki and Biggs, 1999; Kenworthy, 2005). It is, however, an important issue, because successful refugia must not only provide protection from hydraulic forces, but also be associated with bed elements that do not move and crush or dislodge sheltering insects. The relations between hydraulic conditions and bed stability are not simple and require careful consideration. For example, the proportions of the bed that are partially and fully mobile increase with peak discharge (Haschenburger and Wilcock, 2003) suggesting that the extent of stable refugia varies in a fairly simple way during ﬂoods and between ﬂoods of different magnitude. However, such relations might be complicated by the fact that the hydraulic stresses at particular places on a partially mobile bed will respond to transformations of bed texture as a result of bed material deposition and entrainment. Similarly, the bed surrounding a position which offers hydraulic shelter may be mobile so that entrained bed materials may nevertheless impinge on areas that are offering refuge from the ﬂow. There is a fourth factor that is excluded from our experiments, which is likely to be unimportant for this study but may be worthy of future experimental assessment. Animals may migrate down into the hyporheos to avoid or bypass unfavourable hydraulic conditions, although it is likely to be relevant to small-bodied animals and not the large, late-instars of P. latipennis. To date, utilisation of the hyporheos by epigean (surface dwelling) insects has been considered largely in relation to ﬂood disturbances, but it may also be a behaviour that facilitates movement across hydraulically patchy surfaces at more modest ﬂows. Evidence for insect use of the hyporheos in response to hydraulic forcing is equivocal. From observations in a sandy-bottomed channel and experiments conducted in a ﬂume, Palmer et al. (1992) found only very limited evidence for vertical insect movements as water velocity was increased. Working in a fourth-order gravel-bed channel, Olsen and Townsend (2005) found no substantial evidence that invertebrates moved deeper into the hyporheos during ﬂood events. In contrast, Dole-Olivier and Marmonier (1992) found that epigean fauna in the cobble-gravel Miribel Canal do utilise the hyporheos as a refuge during ﬂood events, especially in downwelling zones (Dole-Olivier et al., 1997). The integration of biological factors, sediment transport and the hyporheos in further experiments on insect movement will help to elucidate their relative importance and, ultimately, improve our understanding of ﬂow refugia mechanisms. Continued collaboration between aquatic ecologists and ﬂuvial geomorphologists is central to the success of this endeavour.

Acknowledgements The project was funded by NERC Grant NER/B/S/2000/00697 to Rice, Reid and Lancaster. We are grateful to Natasha Todd-Burley, Mick Barker, David Graham and Stuart Ashby for their help with the ﬂume experiments, which were conducted in

Movements of a macroinvertebrate across a gravel-bed substrate

657

the Department of Civil & Building Engineering at Loughborough University. Ian Atkins and Adam Evans helped with video analysis of insect movements and Jim Chandler provided photogrammetric expertise. We are grateful to two anonymous reviewers for their useful suggestions, which have improved the clarity of the paper.

References Badri, A., Giudicelli, J., Prevot, G., 1987. Effects of ﬂood on the benthic invertebrate community in a Mediterranean river, the Rdat, Morocco. Acta Oecologia Generalis 8, 481–500. Biggs, B.F.J., Duncan, M.J., Francoeur, S.N., Meyer, W.D., 1997. Physical characterisation of microform bed cluster refugia in 12 headwater streams, New Zealand. N. Z. J. Mar. Freshw. Res. 31, 413–422. Bouekaert, F.W., Davis, J., 1998. Microﬂow regimes and the distribution of macroinvertebrates around stream boulders. Freshw. Biol. 40, 77–86. Bufﬁn-Be´langer, T., 2001. Structure d’un e´coulement turbulent dans un cours d’eau a` lit de graviers en pre´sence d’amas de galets. Unpublished Ph.D Thesis, Universite´ de Montre´al, 244pp. Bufﬁn-Be´langer, T., Reid, I., Rice, S., et al., 2003. A casting procedure for reproducing coarse-grained sedimentary surfaces. Earth Surf. Process. Landf. 28, 787–796. Bufﬁn-Be´langer, T., Rice, S.P., Reid, I., Lancaster, J., 2006. Spatial heterogeneity of near-bed hydraulics above a patch of river gravel. Water Resour. Res. 42, W04413, doi:10.1029/2005WR004070. Chandler, J.H., Bufﬁn-Be´langer, T., Rice, S., et al., 2003. The accuracy of simulated riverbed sculpturing system and effectiveness of an amateur digital camera for recording riverbed morphology. Photogramm. Record 18, 209–223. Dole-Olivier, M.-J., Marmonier, P., 1992. Effects of spates on the vertical distribution of the interstitial community. Hydrobiologia 230, 49–61. Dole-Olivier, M.-J., Marmonier, P., Beffy, J.-L., 1997. Response of invertebrates to lotic disturbance: is the hyporheic zone a patchy refugium? Freshw. Biol. 37, 257–276. Drake, T.G., Shreve, R.L., Dietrich, W.E., et al., 1988. Bedload transport of ﬁne gravel observed by motion-picture photography. J. Fluid Mech. 192, 193–217. Finelli, C.M., Hart, D.D., Fonseca, D.M., 1999. Evaluating the spatial resolution of an Acoustic Doppler Velocimeter and the consequences for measuring near-bed ﬂows. Limnol. Oceanogr. 44, 1793–1801. Graham, D.J., Rice, S.P., Reid, I., 2005. A transferable method for the automated grain sizing of river gravels. Water Resour. Res. 41, W07020, doi:10.1029/2004WR003868. Hart, D.D., Clark, B.D., Jasentuliyana, A., 1996. Fine-scale ﬁeld measurement of benthic ﬂow environments inhabited by stream invertebrates. Limnol. Oceanogr. 41, 297–308. Hart, D.D., Finelli, C.M., 1999. Physical–biological coupling in streams: the pervasive effects of ﬂow on benthic organisms. Annu. Rev. Ecol. Syst. 30, 363–395. Hart, D.D., Resh, V.H., 1980. Movement patterns and foraging ecology of a stream caddisﬂy larva. Can. J. Zool. 58, 1174–1185. Haschenburger, J.K., Wilcock, P.R., 2003. Partial transport in a natural gravel-bed channel. Water Resour. Res. 39, 1020, doi:10.1029/2002WR001532. Hershey, A.E., Pastor, J., Peterson, B.J., Kling, G.W., 1993. Stable isotopes resolve the drift paradox for Baetis mayﬂies in an arctic river. Ecology 74, 2315–2325. Holomuzki, J.R., Biggs, B.J.F., 1999. Distributional responses to ﬂow disturbance by a stream-dwelling snail. Oikos 87, 36–47. Huryn, A.D., Denny, M.W., 1997. A biomechanical hypothesis explaining upstream movements by the freshwater snail Elimia. Funct. Ecol. 11, 472–483. Jackson, J.K., McElravy, E.P., Resh, V.H., 1999. Long-term movement of self-marked caddisﬂy larvae, Trichoptera: Sericostomatidae in a California coastal mountain stream. Freshw. Biol. 42, 525–536. Kenworthy, S., 2005. Effects of spatial variability in ﬂow and sediment transport on benthic invertebrates during runoff events: patch and reach scale challenges. EOS Trans. Am. Geophys. Union, 86(18) Joint Assembly Supplement, Abstract B51A-03.

658

S.P. Rice et al.

Kirchner, J.W., Dietrich, W.E., Iseya, F., Ikeda, H., 1990. The variability of critical shear stress, friction angle, and grain protrusion in water-worked sediments. Sedimentology 37, 647–672. Kirkbride, A.D., 1993. Observation of the inﬂuence of bed roughness on turbulence structure in depth limited ﬂows over gravel beds. In: Clifford, N.J., French, J.R., and Hardisty, J. (Eds), Turbulence: Perspectives on Flow and Sediment Transport. Wiley, Chichester, pp. 185–196. Lamarre, H., Roy, A.G., 2005. Reach scale variability of turbulent ﬂow characteristics in a gravel-bed river. Geomorphology 68, 95–113. Lancaster, J., 1999. Small scale movements of lotic macroinvertebrates with variations in ﬂow. Freshw. Biol. 41, 605–619. Lancaster, J., 2000. Geometric scaling of microhabitat patches and their efﬁcacy as refugia during disturbance. J. Anim. Ecol. 69, 442–457. Lancaster, J., Belyea, L.R., 1997. Nested hierarchies and scale-dependence of mechanisms of ﬂow refugium use. J. N. Am. Benthol. Soc. 16, 221–238. Lancaster, J., Belyea, L.R., 2006. Limits to local density: alternative views of abundance–environment relationships. Freshw. Biol. 51, 783–796. Lancaster, J., Bufﬁn-Be´langer, T., Reid, I., Rice, S.P., 2006. Flow- and substratum-mediated movement by a stream insect. Freshw. Biol. 51, 1053–1069. Lancaster, J., Hildrew, A.G., 1993a. Characterizing in-stream ﬂow refugia. Can. J. Fish. Aquat. Sci. 50, 1663–1675. Lancaster, J., Hildrew, A.G., 1993b. Flow refugia and the microdistribution of lotic macroinvertebrates. J. N. Am. Benthol. Soc. 12, 385–393. Lane, S.N., Biron, P.M., Bradbrook, K.F., et al., 1998. Three-dimensional measurement of river channel ﬂow processes using acoustic doppler velocimetry. Earth Surf. Process. Landf. 23, 1247–1267. Lane, S.N., Hardy, R.J., Ingham, D.B., Elliott, L., 2004. Numerical modelling of ﬂow processes over gravelly-surfaces using structured grids and a numerical porosity treatment. Water Resour. Res. 40, W01302, doi:10.1029/2002WR001934. Lawless, M., Robert, A., 2001. Three-dimensional ﬂow structure around small-scale bedforms in a simulated gravel-bed environment. Earth Surf. Process. Landf. 26, 507–522. Matthaei, C.D., Arbuckle, C.J., Townsend, C.R., 2000. Stable surface stones as refugia for invertebrates in a New Zealand stream. J. N. Am. Benthol. Soc. 19, 82–93. Matthaei, C.D., Huber, H., 2002. Microform bed clusters: are they preferred habitats for invertebrates in a ﬂood-prone stream? Freshw. Biol. 47, 2174–2190. Matthaei, C.D., Townsend, C.R., 2000. Inundated ﬂoodplain gravels in a stream with an unstable bed: temporary shelter or true invertebrate refugium? N. Z. J. Mar. Freshw. Res. 34, 147–156. McLelland, S.J., Nicholas, A.P., 2000. A new method for evaluating errors in high frequency ADV measurements. Hydrol. Process. 14, 351–366. Neves, R.J., 1979. Movements of larval and adult Pycnopsyche guttifer, Walker, Trichoptera: Limnephilidae along Factory Brook, Massachusetts. Am. Midl. Nat. 102, 51–58. Nikora, V.I., Goring, D.G., 1998. ADV turbulence measurements: can we improve their interpretation? J. Hydraul. Eng. 124, 630–634. Nikora, V.I., Goring, D.G., Biggs, B.J.F., 1998. On gravel-bed roughness characterisation. Water Resour. Res. 34, 517–527. Olsen, D.A., Townsend, C.R., 2005. Flood effects on invertebrates, sediments and particulate organic matter in the hyporheic zone of a gravel-bed stream. Freshw. Biol. 50, 839–853. Palmer, M.A., Arsenburger, P., Martin, A.P., Denman, D.W., 1996. Disturbance and patch-speciﬁc responses: the interactive effects of woody debris dams on lotic invertebrates. Oecologia 105, 247–257. Palmer, M.A., Bely, A.E., Berg, K.E., 1992. Response of invertebrates to lotic disturbance: a test of the hyporheic refuge hypothesis. Oecologia 89, 182–194. Poff, N.L., Ward, J.V., 1992. Heterogeneous currents and algal resources mediate in situ foraging activity of a mobile stream grazer. Oikos 65, 465–478. Rempel, L.R., Richardson, J.S., Healey, M.C., 1999. Flow refugia for benthic macroinvertebrates during ﬂooding of a large river. J. N. Am. Benthol. Soc. 18, 34–48. Robertson, A.L., Lancaster, J., Belyea, L.R., Hildrew, A.G., 1997. Hydraulic habitat and the assemblage structure of stream benthic microcrustacea. J. N. Am. Benthol. Soc. 16, 562–575.

Movements of a macroinvertebrate across a gravel-bed substrate

659

Smart, G., Aberle, J., Duncan, M., Walsh, J., 2004. Measurement and analysis of alluvial bed roughness. J. Hydraul. Res. 42, 227–237. Statzner, B., Holm, T.F., 1989. Morphological adaptation of shape to ﬂow: microcurrents around lotic macroinvertebrates with known Reynolds numbers at quasi-natural ﬂow conditions. Oecologia 78, 148–157. Sumer, B.M., Chua, L.H.C., Cheng, N.-S., Fredsoe, J., 2003. Inﬂuence of turbulence on bed load sediment transport. J. Hydraul. Eng. 129, 585–596. Vogel, S., 1981. Life in Moving Fluids. Princeton University Press, Princeton, New Jersey, 352pp. Wahl, T., 2000. Analyzing ADV data using WinADV, Joint Conference on Water Resources Engineering and Water Resources Planning & Management, 10pp. Waringer, J.A., 1993. The drag coefﬁcient of cased caddis larvae from running waters: experimental determination and ecological applications. Freshw. Biol. 29, 419–427. Winterbottom, J., Orton, S., Hildrew, A.G., Lancaster, J., 1997. Field experiments on ﬂow refugia in streams. Freshw. Biol. 37, 569–580. Young, W.J., 1992. Clariﬁcation of the criteria used to identify near-bed ﬂow regimes. Freshw. Biol. 28, 383–391.

Discussion by Lynne E. Frostick The authors describe the results of an innovative set of laboratory experiments, which have given some very interesting results. The experiments used a static and impermeable cast of coarse gravel bed which, although realistic in topographic elements, omit consideration of processes linked with hyporheic ﬂows and bed changes during entrainment. These are likely to be most important in unarmoured river beds and in those subject to ‘ﬂashy’ hydrographs. Flume experiments carried out at Hull University into changing bed structure during entrainment have shown that gravels with admixtures of sand dilate during entrainment (Allan and Frostick, 1999; Brasington et al., 2000) and that this results in ﬂows into and out of the bed that have the potential for impacting on the behaviour of any benthic organisms living on and below the bed surface. Have the authors plans for any experiments with less stable beds that might reﬂect these very different conditions? References Allan, A., Frostick, L.E., 1999. Framework dilation, winnowing, and matrix particle size: the behaviour of some sand-gravel mixtures in a laboratory ﬂume. J. Sediment. Res. 69, 21–26. Brasington, J., Frostick, L.E., Middleton, R., Murphy, B.J., 2000. Detecting signiﬁcant sediment motion in a laboratory ﬂume using digital video image analysis. Earth Surf. Process. Landf. 25, 191–196.

Reply by the authors Frostick raises an important issue. Consideration of subsurface spaces to rest, hide and feed is an area of research that could now be pursued, following from the work we have presented here for the relatively simple case of an impermeable bed. The use of a solid cast was beneﬁcial because it allowed us to replicate our experiments (the bed did not change between replications) and provided a means of examining insect

Movements of a macroinvertebrate across a gravel-bed substrate

659

Smart, G., Aberle, J., Duncan, M., Walsh, J., 2004. Measurement and analysis of alluvial bed roughness. J. Hydraul. Res. 42, 227–237. Statzner, B., Holm, T.F., 1989. Morphological adaptation of shape to ﬂow: microcurrents around lotic macroinvertebrates with known Reynolds numbers at quasi-natural ﬂow conditions. Oecologia 78, 148–157. Sumer, B.M., Chua, L.H.C., Cheng, N.-S., Fredsoe, J., 2003. Inﬂuence of turbulence on bed load sediment transport. J. Hydraul. Eng. 129, 585–596. Vogel, S., 1981. Life in Moving Fluids. Princeton University Press, Princeton, New Jersey, 352pp. Wahl, T., 2000. Analyzing ADV data using WinADV, Joint Conference on Water Resources Engineering and Water Resources Planning & Management, 10pp. Waringer, J.A., 1993. The drag coefﬁcient of cased caddis larvae from running waters: experimental determination and ecological applications. Freshw. Biol. 29, 419–427. Winterbottom, J., Orton, S., Hildrew, A.G., Lancaster, J., 1997. Field experiments on ﬂow refugia in streams. Freshw. Biol. 37, 569–580. Young, W.J., 1992. Clariﬁcation of the criteria used to identify near-bed ﬂow regimes. Freshw. Biol. 28, 383–391.

Discussion by Lynne E. Frostick The authors describe the results of an innovative set of laboratory experiments, which have given some very interesting results. The experiments used a static and impermeable cast of coarse gravel bed which, although realistic in topographic elements, omit consideration of processes linked with hyporheic ﬂows and bed changes during entrainment. These are likely to be most important in unarmoured river beds and in those subject to ‘ﬂashy’ hydrographs. Flume experiments carried out at Hull University into changing bed structure during entrainment have shown that gravels with admixtures of sand dilate during entrainment (Allan and Frostick, 1999; Brasington et al., 2000) and that this results in ﬂows into and out of the bed that have the potential for impacting on the behaviour of any benthic organisms living on and below the bed surface. Have the authors plans for any experiments with less stable beds that might reﬂect these very different conditions? References Allan, A., Frostick, L.E., 1999. Framework dilation, winnowing, and matrix particle size: the behaviour of some sand-gravel mixtures in a laboratory ﬂume. J. Sediment. Res. 69, 21–26. Brasington, J., Frostick, L.E., Middleton, R., Murphy, B.J., 2000. Detecting signiﬁcant sediment motion in a laboratory ﬂume using digital video image analysis. Earth Surf. Process. Landf. 25, 191–196.

Reply by the authors Frostick raises an important issue. Consideration of subsurface spaces to rest, hide and feed is an area of research that could now be pursued, following from the work we have presented here for the relatively simple case of an impermeable bed. The use of a solid cast was beneﬁcial because it allowed us to replicate our experiments (the bed did not change between replications) and provided a means of examining insect

Movements of a macroinvertebrate across a gravel-bed substrate

659

Smart, G., Aberle, J., Duncan, M., Walsh, J., 2004. Measurement and analysis of alluvial bed roughness. J. Hydraul. Res. 42, 227–237. Statzner, B., Holm, T.F., 1989. Morphological adaptation of shape to ﬂow: microcurrents around lotic macroinvertebrates with known Reynolds numbers at quasi-natural ﬂow conditions. Oecologia 78, 148–157. Sumer, B.M., Chua, L.H.C., Cheng, N.-S., Fredsoe, J., 2003. Inﬂuence of turbulence on bed load sediment transport. J. Hydraul. Eng. 129, 585–596. Vogel, S., 1981. Life in Moving Fluids. Princeton University Press, Princeton, New Jersey, 352pp. Wahl, T., 2000. Analyzing ADV data using WinADV, Joint Conference on Water Resources Engineering and Water Resources Planning & Management, 10pp. Waringer, J.A., 1993. The drag coefﬁcient of cased caddis larvae from running waters: experimental determination and ecological applications. Freshw. Biol. 29, 419–427. Winterbottom, J., Orton, S., Hildrew, A.G., Lancaster, J., 1997. Field experiments on ﬂow refugia in streams. Freshw. Biol. 37, 569–580. Young, W.J., 1992. Clariﬁcation of the criteria used to identify near-bed ﬂow regimes. Freshw. Biol. 28, 383–391.

Discussion by Lynne E. Frostick The authors describe the results of an innovative set of laboratory experiments, which have given some very interesting results. The experiments used a static and impermeable cast of coarse gravel bed which, although realistic in topographic elements, omit consideration of processes linked with hyporheic ﬂows and bed changes during entrainment. These are likely to be most important in unarmoured river beds and in those subject to ‘ﬂashy’ hydrographs. Flume experiments carried out at Hull University into changing bed structure during entrainment have shown that gravels with admixtures of sand dilate during entrainment (Allan and Frostick, 1999; Brasington et al., 2000) and that this results in ﬂows into and out of the bed that have the potential for impacting on the behaviour of any benthic organisms living on and below the bed surface. Have the authors plans for any experiments with less stable beds that might reﬂect these very different conditions? References Allan, A., Frostick, L.E., 1999. Framework dilation, winnowing, and matrix particle size: the behaviour of some sand-gravel mixtures in a laboratory ﬂume. J. Sediment. Res. 69, 21–26. Brasington, J., Frostick, L.E., Middleton, R., Murphy, B.J., 2000. Detecting signiﬁcant sediment motion in a laboratory ﬂume using digital video image analysis. Earth Surf. Process. Landf. 25, 191–196.

Reply by the authors Frostick raises an important issue. Consideration of subsurface spaces to rest, hide and feed is an area of research that could now be pursued, following from the work we have presented here for the relatively simple case of an impermeable bed. The use of a solid cast was beneﬁcial because it allowed us to replicate our experiments (the bed did not change between replications) and provided a means of examining insect

660

S.P. Rice et al.

movements and the near-bed hydraulics on a facsimile of a natural, water-lain texture and fabric. Nevertheless, we recognise the importance of the subsurface for macroinvertebrates and their movement patterns. There are two main issues. First, ﬂuxes of water into and out of the bed may affect near-bed hydraulics. An impact on local pressure gradients may deﬂect the ﬂow at small scale. Second, subsurface spaces may be utilised and even sought out by larvae, for example, if they are seeking refuge from high ﬂows. We believe that the ﬁrst of these issues is of greater general importance than the second because the consensus of evidence to date suggests that the hyporheos is used by only some organisms some of the time (a detailed argument for this is presented in the penultimate paragraph of our conclusion). Moreover, while the omission of surface–subsurface water exchange may have had an impact on the results we have observed in our experiments, we do not believe that the omission of a hyporheos has had a signiﬁcant impact on our results for two reasons. First, the late instars of P. latipennis that were our subjects are large animals carrying bulky cases which makes it unlikely that they make substantive, if any, use of small subsurface pore spaces characteristic of typical gravel beds in temperate environments (framework gravels with a sandy matrix and a coarse surface layer). Second, the cast is truly three dimensional with excellent reproduction of surface crevices and overhanging particle edges that simulate accurately the surﬁcial interstices of the prototype. In this case, we believe that the incorporation of surface–subsurface water interactions could be an incremental beneﬁt in future experiments. We have considered simulating such exchanges by perforating the cast at points where neighbouring surface ‘particles’ in contact produce interstices. Modifying the cast in this way would produce a very simpliﬁed exchange situation but by continuing to use the ﬁxed-bed cast it would allow for replication and the results could be compared with the data on movements and hydraulics reported in this paper. The question of how insects might respond to dilation and whether, like ﬁne sediments, there is ingress of larvae during dilation events is an intriguing proposition that would require mobile bed experiments to assess and considerable ingenuity to monitor adequately. We suspect that any effect would be highly variable between species and dependent on the traits and morphology of the animals involved. Mobility of framework clasts and interstitial ﬁnes – due to both in situ oscillation as a precursor of entrainment and dilation – makes the channel bed a dangerous place for animals whose body weight is several orders of magnitude smaller than even modestsized mineral grains and who might ﬁnd themselves trapped or crushed. As an avoidance tactic or through reﬂex response to substrate mobility, many animals may drift long before bed dilation occurs.

Gravel-Bed Rivers VI: From Process Understanding to River Restoration H. Habersack, H. Pie´gay, M. Rinaldi, Editors r 2008 Elsevier B.V. All rights reserved.

661

25 Hydraulic geometry of stream reaches and ecological implications Nicolas Lamouroux

Abstract The hydraulic geometry (width- and depth-discharge relationships) of stream reaches (lengthten times the width) is a key physical description for predicting the ecological impact of physical constraints at a range of spatial scales. Downstream and at-a-reach hydraulic geometry relationships make it possible to estimate reach-averaged hydraulic variables across reaches and within reaches when discharge rate changes. Reach-averaged hydraulics govern the frequency distribution of point hydraulic variables (velocity, shear stress, water depth) in stream reaches. Therefore, the hydraulic geometry of reaches is linked to the diversity of microhabitat conditions, which constrain aquatic organisms. As a consequence, reach-scale responses of populations and communities are also related to the hydraulic geometry. Some of these relationships are very general; for example a number of functional descriptions of ﬁsh and macroinvertebrate communities (describing size, shape, behaviour, reproduction patterns) show similar responses to hydraulics in stream reaches of independent basins or continents. The hydraulic geometry of reaches provides a tool for predicting the ecological impacts of stream restoration (discharge modiﬁcations, other hydraulic changes) over large scales. An example of application is the estimation of suitable ﬂow rates for populations over whole river networks. 1.

Introduction

Interactions between physical and biological processes in streams are fascinating in many ways. First, both physical and biological processes occur over a hierarchy of spatial and temporal scales (Frissell et al., 1986; Biggs et al., 2005). Interactions occur at all scales. Taking somewhat extreme examples, local turbulence patterns affect the swimming performance and behaviour of ﬁsh individuals (Nikora et al., 2003). In return, freshwater organisms modify the physical properties of their microhabitat such as the critical shear stress of bed particles (Statzner et al., 2003). At much larger E-mail address: [email protected] ISSN: 0928-2025

DOI: 10.1016/S0928-2025(07)11153-6

N. Lamouroux

662

scales, differences in ﬁsh communities can be related to distinct evolutionary histories. Over time, environmental constraints played a large role in shaping distinct ﬁsh communities across regions and continents (Moyle and Herbold, 1987; Montgomery, 2004). Second, the physical and biological processes occurring in streams are complex in nature. For example bursts, hydraulic jumps, secondary currents are common features in natural rivers. As an outcome of this complexity, observed physical patterns (e.g., velocity proﬁles, friction, bed particle stability, turbulence) often differ from theoretical expectations assuming more uniform and stationary ﬂow conditions (Bufﬁn-Be´langer et al., 2000). Management issues certainly challenge our ability to cope with this complexity for modelling the ecological effects of ﬂow changes and morphological modiﬁcations in rivers. Stream management naturally evolves towards a more integrated vision of stream networks, at the basin or larger spatial scales, where water policies are deﬁned and the effects of large-scale environmental constraints (e.g., ﬂow anomalies resulting from climate change) have to be assessed. This evolution towards large-scale management creates an additional challenge. It stresses the need for quantitative models of physical and biological processes that are as ‘general’ as possible, and transferable across streams of whole river networks. For many stream management issues, available ‘general’ knowledge is often qualitative rather than quantitative. For example, the ecological management of ﬂow regimes in streams relies more on qualitative lessons assembled from many case studies than on quantitative tools applicable across stream networks (Poff et al., 1997). Research is needed to identify the whole range of scales at which interactions between physical and biological processes occur (Hart and Finelli, 1999; Biggs et al., 2005), and for identifying key processes and key scales for modelling the functioning of stream ecosystems. An important task in this research is to better understand the links between processes occurring at different spatial and temporal scales. In this paper, I gather a number of results from studies that support the idea that the hydraulic geometry of stream reaches (depth- and width-discharge relationships) is a key ecohydraulic description of streams that contributes to integrate knowledge and models developed at various spatial scales, ranging from the microhabitat to the catchment. After brieﬂy introducing hydraulic geometry relationships and their properties, I describe how hydraulic geometry is linked to the distribution of microhabitat hydraulics within reaches. Then, I show that the outputs of conventional instream habitat models (linking hydraulic models of reaches to biological models of preference for microhabitat hydraulics) are predictable from the reach-averaged hydraulics. I also give examples of relationships between hydraulics and actual stream communities that were found to be consistent at the microhabitat and stream reach scales. I ﬁnally discuss how the properties of hydraulic geometry relationships across river networks contribute to the understanding and management of stream communities at large spatial scales.

2.

The hydraulic geometry of stream reaches

At-a-station hydraulic geometry refers to the way in which the ﬂow geometry (depth, wetted width, velocity) changes at a single cross-section as discharge varies through

Hydraulic geometry of stream reaches and ecological implications

663

time. Downstream hydraulic geometry (i.e., regime equations) refers to the way in which the channel geometry changes along or between rivers due to longitudinal increases in a given ﬂow discharge statistic (e.g., mean ﬂow or bankfull ﬂow). These terms were introduced by Leopold and Maddock (1953), who suggested that hydraulic geometry relationships could be ﬁtted to power laws as W ¼ aQb

(25.1)

D ¼ cQf

(25.2)

for the at-a-station hydraulic geometry, where W and D are water width and depth at instantaneous discharge Q, and ¯ bd ¯ ¼ ad Q W

(25.3)

¯ fd ¯ ¼ cd Q D

(25.4) ¯ ¯ for the downstream hydraulic geometry, where W and D are water width and depth at ¯ parameters a, b, c, f and ad, bd, cd, fd the hydraulic geometry mean discharge Q; coefﬁcients and exponents. Hydraulic geometry relationships generally involve a third equation reﬂecting ata-station or downstream changes in velocity, and sometimes others involving, for example slope and friction. Equations for velocity will be omitted here considering that mass conservation implies V¼

Q DW

(25.5)

¯ Q (25.6) V¯ ¼ ¯W ¯ D ¯ Similarly, equations for depth and width can where V and V¯ are velocities at Q and Q. be used to generate equations for other hydraulic variables such as the dimensionless Froude number and Reynolds number, which will be used in this paper as Fr ¼

Q g0:5 D1:5 W

(25.7)

Q (25.8) nW ¯ will be noted as Fr and Re. Re is equivalent to a speciﬁc discharge if we Fr and Re at Q assume the water kinematic viscosity n to be constant; g is the acceleration due to gravity. Coefﬁcients and exponents of hydraulic geometry relationships share similar ranges in different geomorphologic contexts (e.g., Park, 1977; Knighton, 1998). Typical average values for the exponents found in the literature are around 0.15 for b, 0.4 for f, 0.5 for bd and 0.36 for fd (Knighton, 1998). Therefore, hydraulic geometry relationships have been widely used as empirical ‘translators’ between information on discharge and information on hydraulics and channel form (Ferguson, 1986; Rhoads, 1994), for example for predicting stable channel forms for given discharge rates, for Re ¼

N. Lamouroux

664

predicting past discharge rates from actual channel forms or for extrapolating hydraulic properties and transfer rates across discharge rates. Analytical derivations of hydraulic geometry relationships have been developed (Parker, 1978; Ferguson, 1986; Grifﬁths, 2003). The variations in exponents and coefﬁcients of hydraulic geometry relationships across sites remain largely unexplained. Concerning at-a-station relationships, exponents b and f vary among channel types (straight, meandering, braided), between sections in pools and sections in rifﬂes, with bank stability and channel boundary composition (Ferguson, 1986; Millar and Quick, 1993; Knighton, 1998). For reaches at quasi-equilibrium, variations in at-a-station coefﬁcients a and c are addressed by studies of downstream hydraulic geometry, because at-a-station hydraulic geometry relationships (equations (25.1) and (25.2)) can be re-written as b Q ¯ (25.9) W ¼W ¯ Q f Q ¯ D¼D ¯ Q

(25.10)

For downstream relationships, some variation in coefﬁcients and exponents has been explained by, for example bank vegetation, channel sediment characteristics and ﬂood magnitude (Rhoads, 1991). Overall, as for at-a-station exponents, variations in parameters of downstream hydraulic geometry are probably more related to local channel factors than regional ones (Ridenour, 2001). Though generally developed for cross-sections, deﬁning hydraulic geometry for stream reaches (length ten times the stream width) is attractive for a number of purposes. Flow resistance, channel forms and hydraulic conditions may be better understood at the reach scale, where the variability among cross-sections is smoothed. Hydraulic geometry of reaches (i.e., involving reach-averaged depth and width) should show less variability across sites than the hydraulic geometry of cross-sections, which is sensitive to the choice of these sections (Knighton, 1998; Stewardson, 2005). Following Jowett (1998) and Lamouroux and Capra (2002), I will refer implicitly to the hydraulic geometry of reaches (and not cross-sections) in the following section.

3.

Hydraulic geometry of reaches and microhabitat hydraulics

The microhabitat scale (organism’s scale, e.g., 1 m2 for ﬁsh) is a major scale for studies of biological responses to hydraulic constraints. Between microhabitats within reaches, ﬁsh, invertebrates, plants and amphibians show marked preferences for distributions of local sets of point velocities, water depths or bottom shear stresses (e.g., Gore and Judy, 1981; Bovee, 1982; Kupferberg, 1996). Hydraulic preferences of species at the microhabitat scale are certainly ﬂexible across streams, seasons or life stages (Leftwich et al., 1997); they also vary with discharge rate, habitat availability and daytime. However, many taxa show consistent hydraulic preferences across streams in a regional context (Lamouroux et al., 1999). Beyond species preferences for hydraulics,

Hydraulic geometry of stream reaches and ecological implications

665

traits of stream assemblages reﬂecting, for example their morphology, behaviour, longevity, feeding and reproductive strategy show expected responses to microhabitat hydraulics. These responses can be interpreted in terms of drag reduction and life history trade-offs (Sagnes et al., 1997; Lamouroux et al., 2004; Puijalon and Bornette, 2004). Models based on species trait responses to hydraulics are potentially less constrained by biogeography and the taxonomic context than species-speciﬁc models. It is easier to interpret the relationships between stream assemblages and stream hydraulics at the microhabitat scale than at the reach or larger scales. At the reach scale, the mere existence of downstream hydraulic geometry relationships illustrates the inter-correlation of physical variables, particularly along longitudinal gradients. Temperature, water quality and many other variables also vary along longitudinal gradients. As a result, it is difﬁcult to determine which reach-scale physical variables are responsible for the biological responses. By contrast, within a reach, differences in species density across microhabitats are rarely due to temperature or water quality variations, and cannot reﬂect (though they can be affected by) the effects of migration barriers or biogeography. They are conﬁdently interpreted as being related to hydraulic variation. The importance of microhabitat hydraulics for stream communities emphasises the need to describe them as well as their variations with time (discharge) and with stream management options (changes in ﬂow regime and morphology). Traditional approaches of hydraulic engineers have naturally been used for this purpose (Fig. 25.1). However, deterministic models providing numeric solutions of the Navier–Stokes equations or simpliﬁcations of them (e.g., Bovee, 1982; Guay et al., 2000) are generally little adapted to modelling microhabitat hydraulics of complex natural ﬂows (e.g., Osborne et al., 1988). In intermediate- to large-scale roughness streams or under low ﬂow conditions, hydraulic jumps and local energy losses are the common rule. Estimates of point velocity with relative errors of 100% are common in such situations (Guay et al., 2001). Dingman (1989) proposed a parametric formulation of the point velocity frequency distributions in cross-sections. His approach contrasted with others by looking at the distributions of point velocity in cross-sections from a statistical, descriptive point of view. However, he could not relate the distribution parameters to any explanatory variable. Extending his approach at the scale of stream reaches, Lamouroux et al. (1992, 1995) and Stewardson and MacMahon (2002) showed that the probability distributions of point bed shear stress, point velocity and water depth in stream reaches share common shapes across stream reaches. These shapes vary similarly as a function of discharge rate within reaches. Moreover, these probability distributions are largely predictable, across a wide range of stream reaches, from the at-a-reach hydraulic geometry (Fig. 25.2). The distributions of point velocity and bottom shear stress in reaches (as estimated from the alternative method of Statzner and Mu¨ller, 1989) vary from a bimodal distribution at low ﬂow rates towards a normal distribution at high ﬂow rates (Fig. 25.2). These properties are shared by a very wide range of stream reaches (from mountain to ﬂoodplain channels; with sand to boulders) in different regions and countries, though the velocity distributions were studied in a wider range of reaches (width 5–110 m, particle size up to 50 cm in Lamouroux et al., 1995) than the shear stress distributions (width 1–30 m, particle size up to 3 cm in Lamouroux et al., 1992). Knowledge of the

666

N. Lamouroux

Figure 25.1. Schematic principle of conventional instream habitat models. A hydraulic model (A) predicts the values of point hydraulic variables (water column velocity and depth calculated for various discharge rates) in individual cells within a stream reach. Suitability curves (B) indicate the preference (generally scores between 0 and 1) of any species for point hydraulic variables. They are used to transform point hydraulic variables (A) into point habitat values (HV) in each cell (C). A reach HV (ranging between 0 and 1) is a weighted average of point HVs. Its variations with discharge (D) are used to estimate the impact of ﬂow management on the habitat of the species. Drawn using EVHA software (Ginot, 1998).

at-a-reach depth- and width-discharge relationships are sufﬁcient to predict those distributions with reasonable accuracy. For example, 70% of the variability in frequency of low point velocities (ohalf of the reach-averaged velocity) across streams is explained by the at-a-reach hydraulic geometry. The distributions of point depths are less variable than those of velocities or shear stresses, but are more difﬁcult to predict directly from the hydraulic geometry of reaches (Lamouroux, 1998; Stewardson and McMahon, 2002). Nevertheless, their variations with discharge rates are also related to the at-a-reach hydraulic geometry in a wide range of streams (width 1–300 m, particle size up to 50 cm in Lamouroux, 1998). Statistical models of microhabitat hydraulics share a number of attractive and limiting properties. Among positive aspects, they are general and demonstrate the possibility of using the at-a-reach hydraulic geometry to estimate microhabitat hydraulics at different discharge rates. In particular, the Froude number of reaches, Fr, is a key variable governing the shape of the distributions of all microhabitat variables. Using statistical hydraulic models requires few ﬁeld measurements (i.e., the estimation of the at-a-reach hydraulic geometry), little experience, no velocity measurements and no spatially explicit topography. Among their limits, statistical models

Hydraulic geometry of stream reaches and ecological implications

667

Figure 25.2. Examples of frequency distributions within stream reaches, shown at three discharge rates, of the relative point velocity (point velocity divided by its mean reach value), the bed shear stress (estimated from FST-hemisphere numbers, Statzner and Mu¨ller, 1989) and the relative point depth (shown for a small German brook and large French river). Bars correspond to observed distributions, lines correspond to modelled distributions (using statistical hydraulic models) based on the at-a-reach hydraulic geometry. These examples are drawn using data and models in Lamouroux et al. (1992, 1995) and Lamouroux (1998).

apply only in streams with morphology close to natural. They are not adapted, for example to fully channelised rivers where velocity and depth distribution would be more homogeneous than those shown in Fig. 25.2. Contrary to deterministic approaches, statistical models are not adapted to artiﬁcial situations or for describing

668

N. Lamouroux

spatially explicit hydraulic patterns of potential importance for aquatic organisms (e.g., secondary currents, hydraulic refuges near banks or below woody debris). Finally, their application in particular morphologies such as those of braided rivers certainly requires further testing. Statistical approaches could be useful for describing a number of complex hydraulic patterns (e.g., turbulence intensity, frequency of hydraulic jumps, frequency of disturbance of mixed sediment patches) that are difﬁcult to predict using deterministic approaches. Beyond making a link between the hydraulic geometry of reaches and microhabitat-scale hydraulic diversity, statistical approaches also revealed that this link does not follow simple linear forms. The examples of Fig. 25.2 illustrate that for a given reach-averaged velocity, speciﬁc proportions of low ﬂow velocities and patches of high velocity both occur in a natural reach. At low ﬂow, the reach-averaged velocity or shear stress may occur infrequently in the reach, potentially explaining our difﬁculties to apply conventional approaches of hydraulic engineers in natural reaches. The diversity of microhabitat hydraulics within reaches is of importance for the dislodgement of particles and organisms, the retention of nutrients and pollutants, spawning activities and many other processes. The heterogeneous forms of velocity and shear stress distributions at low ﬂows probably result from hydraulic variability along three dimensions, that is the longitudinal dimension along which sequences of various geomorphologic units are organised, the transversal dimension along which preferential ﬂow paths contrast with ﬂows near the bank and the vertical dimension along which bed particles create hydraulic shelters.

4.

Hydraulic geometry of reaches and reach habitat values for aquatic species

The need to quantify the impacts of hydrological (changes in discharge regime, minimum ﬂows, ﬂooding properties) or morphological (e.g., channelisation or weir construction) modiﬁcations of rivers led to the development of instream habitat models linking hydraulic models of stream reaches and models of biological responses to microhabitat hydraulics (Fig. 25.1). Instream habitat models became popular and resulted in a wealth of instream ﬂow studies in the last two decades (Reiser et al., 1989; Guay et al., 2000). Conventional instream habitat models (e.g., the physical habitat simulation system (PHABSIM) – Bovee, 1982) link a traditional hydraulic engineering model to habitat suitability curves for water depth, point velocity, bed particle size and other microhabitat variables (e.g., cover). The hydraulic model predicts the values of point habitat variables (water column velocity, depth) as a function of the discharge in individual cells within a stream reach (Fig. 25.1A). Suitability curves (Fig 25.1B) are used to transform point habitat variables into point habitat values (HV scores ranging between 0 and 1) in each cell (Fig. 25.1C). A ‘weighted usable area’ (WUA) is computed at the reach scale as the sum (across cells) of point HVs multiplied by the area of the cell. A reach HV (ranging between 0 and 1) can be calculated as WUA divided by the wetted area of the reach (Fig. 25.1D). Therefore, the major reach-scale outputs of conventional instream habitat models are HV or WUA changes with discharge rate (Fig. 25.1D).

Hydraulic geometry of stream reaches and ecological implications

669

The basic principle of instream habitat models is to describe microhabitat hydraulics within stream reaches and link them to models of species preferences for hydraulics. Because microhabitat hydraulics are largely predictable from the at-a-reach hydraulic geometry (see previous part), results of instream habitat models should also depend on the hydraulic geometry of reaches. Sensitivity analyses of conventional instream habitat models calibrated in 200 stream reaches of France and New Zealand (width 4–140 m) conﬁrmed these expectations (Fig. 25.3). When the results of instream habitat models for a given microhabitat preference model (i.e., a given taxa of ﬁsh or invertebrate) were compared across stream reaches and discharge rates, between 54 and 95% of the variation in reach HV was explained by the knowledge of depth- and width-discharge relationships (and average particle size), depending on the taxa considered. This made it possible to derive generalised instream habitat models, whose input variables are the at-a-reach hydraulic geometry relationships, as simple alternatives to conventional habitat models. Tests of generalised models calibrated in France were very satisfactory when applied in New Zealand, and most models developed in New Zealand gave reasonable accuracy when applied in rivers larger or smaller than those used to calibrate them. This suggests that generalised habitat models based on the at-a-reach hydraulic geometry have the potential to be very general, that is applicable in many places in the world. The sensitivity analysis of instream habitat models revealed key physical variables that are expected to drive HVs in stream reaches. The discharge per unit width (or speciﬁc discharge of reaches) governs the taxa-speciﬁc shape of HV changes within reaches, in France and New Zealand (Fig. 25.3). This speciﬁc discharge may also be

Figure 25.3. Examples of reach habitat values as a function of speciﬁc discharge in four reaches of France (for two ﬁsh species, left columns) and New Zealand (for a mayﬂy, right column). Dots correspond to habitat values estimated by conventional instream habitat models (Fig. 25.1), lines correspond to habitat values estimates (using generalised habitat models) based on the at-a-reach hydraulic geometry. These examples are drawn using data and models in Lamouroux and Capra (2002) and Lamouroux and Jowett (2005).

N. Lamouroux

670

quantiﬁed as a Reynolds number of reaches, Re, if we consider water viscosity n as a constant. For example, the generalised models suggested that many taxa have a maximum habitat suitability for a speciﬁc discharge around Q/W ¼ 0.3 m2/s in most reaches of New Zealand (Lamouroux and Jowett, 2005). Such conditions may optimise the energy balance (drag vs. food drift) for a number of species in New Zealand. The mean Froude number of reaches Fr was generally the major variable explaining overall HV differences between reaches, independently of discharge rate, both in France and New Zealand. Fr reﬂects the proportion of rifﬂes versus pools in reaches because point Fr has contrasted values in these habitat types (Jowett, 1993). Therefore, the general inﬂuence of Fr on HVs conﬁrms the importance of channel longitudinal morphology for the habitat suitability of freshwater populations (Niemi et al., 1990; Nelson et al., 1992). Hydraulic geometry was considered as a potential predictor of habitat suitability of stream reaches in a number of earlier studies (Bovee, 1982; Mosley, 1983; Hogan and Church, 1989). In most approaches, the hydraulic geometry of reaches was used to describe the average hydraulic characteristics of reaches; then, simpliﬁed procedures were used to estimate microhabitat hydraulics and link them to biological preference models. For example, Jowett (1998) considered point velocities in reaches as equal to their average at the reach scale, while others assumed that point velocities had a given probability distribution (Singh and McConkey Broeren, 1989; Rao et al., 1993). The generalised habitat models described above go somewhat further as they do not require particular assumptions on the statistical distribution of point hydraulic variables. Altogether, these studies reﬂect the potential of hydraulic geometry relationships of reaches for estimating habitat suitability for various species.

5.

Hydraulic geometry of reaches and actual communities

The extent to which microhabitat preference of species inﬂuence community structure at larger scales is still largely unknown. Biological validations of instream habitat models are mainly limited to salmonids in small streams and are often site-speciﬁc (Orth and Maughan, 1982; Jowett, 1992; Phillips et al., 2000). Lacking clear validation, instream habitat models have frequently been criticised or defended (Orth and Maughan, 1982), leaving managers in uncertainty. Two recent studies (Lamouroux et al., 2002, 2004) that dealt with ﬁsh and macroinvertebrates showed, however, that microhabitat preferences within reaches are consistent with variation in reach-scale community patterns observed in independent regions or continents. Both studies involved numerous reaches across France (invertebrates) or France and North America (ﬁsh). In both studies, reach-scale communities were described by proportions of individuals sharing particular species traits (morphology, behaviour, feeding habitats, reproductive strategy, longevity and others). Reach-scale hydraulics were described by several variables including the Froude number Fr at low ﬂow conditions. Fig. 25.4 provides examples of results obtained in these studies. Within French reaches, microhabitats with high Froude numbers tend to contain fewer fecund ﬁsh (Fig. 25.4A, B). This relationship is highly scattered at the microhabitat scale.

Hydraulic geometry of stream reaches and ecological implications

671

Figure 25.4. Examples of consistent responses of ﬁsh fecundity (proportion of individual ﬁsh belonging to species spawning more than 100,000 eggs/yr) to hydraulics at the microhabitat scale (within reaches) and at the reach scale (within France and Virginia). Linear regressions are given for illustration and comparing effects. Within French reaches, microhabitats with high Froude numbers tend to contain little fecund ﬁsh: (A) Ain River in July, 1991, r2 ¼ 0.12, P ¼ 0.02, slope ¼ 145; (B) Arde`che River in April, 1995, r2 ¼ 0.17, P ¼ 0.01, slope ¼ 243. Reaches with high Fr (at minimum monthly ﬂow) contain little fecund ﬁsh, both (C) in France (r2 ¼ 0.44, P ¼ 0.03, slope ¼ 131) and (D) in Virginia (r2 ¼ 0.43, P ¼ 0.05, slope ¼ 169). These examples are selected among several in Lamouroux et al. (2002).

However, it reﬂects a trend that is consistently repeated within rivers and, over time, probably explains the reach-scale response of ﬁsh fecundity to hydraulics shown in Fig. 25.4C, D. More generally, these studies suggested that fast-ﬂowing and shallow reaches associated with high Fr at low ﬂows generally contain higher proportions of streamlined and small ﬁsh and invertebrates, suggesting morphological adaptations for reducing drag in these conditions (Sagnes et al., 1997). Communities in these reaches are more short-lived than others and less fecund. Invertebrates in reaches with high Fr have attachment systems and foraging strategies adapted to habitats with high shear stresses. Such responses to hydraulics can be very general.

N. Lamouroux

672 6.

Discussion: hydraulic geometry as a tool for large-scale ecological modelling

Previous sections demonstrated that the hydraulic geometry of reaches show general patterns across reaches. The at-a-reach relationships are also generally linked, in a wide range of streams, (1) to microhabitat hydraulics, (2) to the output of conventional instream habitat models and (3) to characteristics of actual stream communities. In this context, models of hydraulic geometry are particularly attractive to ecologists. They make it possible to model the ecological impacts of discharge management across river networks. Let us consider a simple example. By running instream habitat models in a number of streams in New Zealand, Lamouroux and Jowett (2005) showed that for many freshwater species in this country, the suitability of hydraulic habitat reaches a maximum when Q a ¼ 0:3 m2 s1 W

(25.11)

This information basically reﬂects a ‘global’ result obtained from studies of microhabitat preferences of stream organisms within reaches. Assuming that at-a-reach hydraulic geometry relationships follow equations (25.9) and (25.10), and that downstream hydraulic geometry relationships follow equations (25.3) and (25.4), the discharge rate corresponding to maximum suitability can be estimated in any reach as h i1=ð1bÞ ¯ ðbd bÞ Q ¼ aad Q

(25.12)

A general equation, such as equation (25.12), can help deﬁne policies for minimum ﬂows at large-scales. It shows that the discharge rate corresponding to optimum habitat suitability should, on average across reaches, depend on the parameters of ata-reach (b) and downstream (ad and bd) hydraulic geometry relationships appropriate for describing these reaches (e.g., Jowett, 1998). Equation (25.12) reﬂects a scaled-up expression of microhabitat preferences of aquatic organisms for estimating suitable discharge rates across reaches, at the catchment or larger scales. It can also be used for reﬂecting the ecological impact of different hydraulic geometry parameters in different catchments. In the same vein, Lamouroux and Souchon (2002) linked generalised instream habitat models to hydraulic geometry relationships for French reaches to estimate the sensitivity of ﬁsh HVs to discharge changes. Such an exercise is attractive in the context of a global climate change. It makes it possible to identify types of streams where expected discharge anomalies should strongly modify instream HVs. General equations such as equation (25.12) are attractive but should be used and interpreted with care. Equation (25.12) results from the coupling of various models (hydraulic geometry relationships, generalised instream habitat models, microhabitat suitability models) whose assumptions and limits should always be kept in mind. For example, depth- and width-discharge relationships in reaches do not always follow power laws; a high degree of variation in hydraulic geometry exponents and coefﬁcients remains unexplained; generalised instream habitat models have outputs that can deviate from those of conventional habitat models in some streams; generalised instream habitat models apply to reaches with a morphology close to natural;

Hydraulic geometry of stream reaches and ecological implications

673

microhabitat preferences of species are ﬂexible across streams and life stages (see previous sections). Trying to ﬁnd an average model relating hydraulics and ﬁsh communities and trying to understand variability around this average model are equally useful. When deriving general models for the management of river networks, we necessarily lose precision and make a number of assumptions. Equation (25.12) is certainly not relevant in some stream reaches in New Zealand, for example those where species microhabitat preferences deviate from the average available model, or those where the physical habitat is not a limiting factor for aquatic populations. Nevertheless, equation (25.12) is probably relevant when expressed for groups of streams, for example when comparing suitable discharge rates in small versus large streams or between two catchments. Overall, such equations should be used with estimates of their uncertainties. Uncertainties for hydraulic geometry coefﬁcients are possible to estimate, given the range of values available in the literature (Jowett, 1998). Uncertainties for the value of a are available from tests of generalised instream habitat models (Lamouroux and Jowett, 2005). Uncertainties for the actual response of aquatic communities are also becoming available (Jowett and Biggs, 2004). Associated with their uncertainties and limits, models of biological response to hydraulics at large scales are useful for both the management of rivers and our understanding of the relations between physical and biological processes that are observed at various scales. Studies of the hydraulic geometry of reaches should continue to play an important role in this research ﬁeld. Hopefully, the ecological implications of hydraulic geometry will stimulate a stronger collaboration between ecologists, geomorphologists and hydraulic engineers.

Acknowledgements Many thanks to Ton Snelder, Martin Doyle, Jason Julian, Rob Ferguson, Massimo Rinaldi and an anonymous reviewer for their helpful comments.

References Biggs, B.J.F., Nikora, V.I., Snelder, T.H., 2005. Linking scales of ﬂow variability to lotic ecosystem structure and function. River Res. Appl. 21, 283–298. Bovee, K.D., 1982. A guide to stream habitat analysis using the instream ﬂow incremental methodology. Instream Flow Information Paper 12. U.S. Fish and Wildlife Service, Fort Collins, CO. Bufﬁn-Be´langer, T., Roy, A.G., Kirkbride, A.D., 2000. On large-scale ﬂow structures in a gravel-bed river. Geomorphology 32, 417–435. Dingman, S.L., 1989. Probability distribution of velocity in natural cross sections. Water Resour. Res. 25, 509–518. Ferguson, R.I., 1986. Hydraulics and hydraulic geometry. Prog. Phys. Geogr. 10, 1–31. Frissell, C.A., Liss, W.J., Warren, C.E., Hurley, M.D., 1986. A hierarchical framework for stream habitat classiﬁcation: viewing streams in a watershed context. Environ. Manage. 10, 199–214. Ginot, V., 1998. Logiciel EVHA. Evaluation de l’habitat physique des poissons en rivie`re (version 2.0). Cemagref Lyon et Ministe`re de l’ame´nagement du Territoire et de l’Environnement, Direction de l’Eau, Paris, France.

674

N. Lamouroux

Gore, J.A., Judy, R.D., 1981. Predictive models of benthic macroinvertebrate density for use in instream ﬂow and regulated ﬂow management. Can. J. Fish. Aquat. Sci. 38, 1363–1370. Grifﬁths, G.A., 2003. Downstream hydraulic geometry and hydraulic similitude. Water Resour. Res. 39, 1094–1099. Guay, J.C., Boisclair, D., Leclerc, M., et al., 2001. Science on the edge of spatial scales: a reply to the comments of Williams. Can. J. Fish. Aquat. Sci. 58, 2108–2111. Guay, J.C., Boisclair, D., Rioux, D., et al., 2000. Development and validation of numerical habitat models for juveniles of Atlantic salmon (Salmo salar). Can. J. Fish. Aquat. Sci. 57, 2065–2075. Hart, D.D., Finelli, C.M., 1999. Physical-biological coupling in streams: the pervasive effects of ﬂow on benthic organisms. Annu. Rev. Ecol. Syst. 30, 363–395. Hogan, D.L., Church, M., 1989. Hydraulic geometry in small, coastal streams: progress towards quantiﬁcation of salmonid habitat. Can. J. Fish. Aquat. Sci. 46, 844–852. Jowett, I.G., 1992. Models of the abundance of large brown trout in New Zealand rivers. North Am. J. Fish. Manage. 12, 417–432. Jowett, I.G., 1993. A method for objectively identifying pool, run, and rifﬂe habitats from physical measurements. New Zeal. J. Mar. Freshwat. Res. 27, 241–248. Jowett, I.G., 1998. Hydraulic geometry of New Zealand rivers and its use as a preliminary method of habitat assessment. Regul. Rivers Res. Manage. 14, 451–466. Jowett, I.G. and Biggs, B.J.F., 2004. The effectiveness of habitat-based minimum ﬂow assessments – six case studies, Proceedings of the Fifth International Symposium on Ecohydraulics, September 12–17, IAHR, Madrid, pp. 695–701. Knighton, D., 1998. Fluvial forms and processes: a new perspective. Wiley, New York. Kupferberg, S.J., 1996. Hydrologic and geomorphic factors affecting conservation of a river-breeding frog (Rana boylii). Ecol. Appl. 6, 1332–1344. Lamouroux, N., 1998. Depth probability distributions in stream reaches. J. Hydraul. Eng. 124, 224–227. Lamouroux, N., Capra, H., 2002. Simple predictions of instream habitat model outputs for target ﬁsh populations. Freshwat. Biol. 47, 1543–1556. Lamouroux, N., Capra, H., Pouilly, M., Souchon, Y., 1999. Fish habitat preferences at the local scale in large streams of southern France. Freshwat. Biol. 42, 673–687. Lamouroux, N., Dole´dec, S., Gayraud, S., 2004. Biological traits of stream macroinvertebrate assemblages: effects of microhabitat, reach and basin ﬁlters. J. North Am. Benthol. Soc. 23, 449–466. Lamouroux, N., Jowett, I.G., 2005. Generalized instream habitat models. Can. J. Fish. Aquat. Sci. 62, 7–14. Lamouroux, N., Poff, N.L., Angermeier, P.L., 2002. Intercontinental convergence of stream ﬁsh community traits along geomorphic and hydraulic gradients. Ecology 83, 1792–1807. Lamouroux, N., Souchon, Y., 2002. Simple predictions of instream habitat model outputs for ﬁsh habitat guilds in large streams. Freshwat. Biol. 47, 1531–1542. Lamouroux, N., Souchon, Y., He´rouin, E., 1995. Predicting velocity frequency distributions in stream reaches. Water Resour. Res. 31, 2367–2375. Lamouroux, N., Statzner, B., Fuchs, U., et al., 1992. An unconventional approach to modeling spatial and temporal variability of local shear stress in stream segments. Water Resour. Res. 28, 3251–3258. Leftwich, K.N., Angermeier, P.L., Dolloff, C.A., 1997. Factors inﬂuencing behavior and transferability of habitat models for a benthic stream ﬁsh. Trans. Am. Fish. Soc. 126, 725–734. Leopold, L.B. and Maddock, T., 1953. The hydraulic geometry of stream channels and some physiographic implications. U.S. Geological Survey, Professional Paper 252. Washington, DC. Millar, R.G., Quick, M.C., 1993. Effect of bank stability on geometry of gravel rivers. J. Hydraul. Eng. 119, 1343–1363. Montgomery, D.R., 2004. Geology, geomorphology, and the restoration ecology of salmon. GSA Today 14, 4–12. Mosley, M.P., 1983. Response of braided rivers to changing discharge. J. Hydrol. 22, 18–63. Moyle, P.B., Herbold, B., 1987. Life-history patterns and community structure in stream ﬁshes of western North America: comparisons with eastern North America and Europe. In: Matthews, W.J. and Heins, D.C. (Eds), Community and Evolutionary Ecology of North American Stream Fishes. University of Oklahoma Press, Oklahoma, pp. 25–32.

Hydraulic geometry of stream reaches and ecological implications

675

Nelson, R.L., Platts, W.S., Larsen, D.P., Jensen, S.E., 1992. Trout distribution and habitat in relation to geology and geomorphology in the North Fork Humboldt River drainage, northeastern Nevada. Trans. Am. Fish. Soc. 121, 405–426. Niemi, G.J., De Vore, P., Detenbeck, N., et al., 1990. Overview of case studies on recovery of aquatic systems from disturbance. Environ. Manage. 14, 571–587. Nikora, V.I., Aberle, J., Biggs, B.J.F., et al., 2003. Effects of ﬁsh size, time-to-fatigue and turbulence on swimming performance: a case study of Galaxias maculatus. J. Fish Biol. 63, 1365–1382. Orth, D.J., Maughan, O.E., 1982. Evaluation of the incremental methodology for recommending instream ﬂow for ﬁshes. Trans. Am. Fish. Soc. 111, 413–445. Osborne, L.L., Wiley, M.J., Larimore, R.W., 1988. Assessment of the water surface proﬁle model: accuracy of predicted instream ﬁsh habitat conditions in low-gradient, warmwater streams. Regul. Rivers Res. Manage. 2, 619–631. Park, C.C., 1977. World-wide variations in hydraulic geometry exponents of stream channels: an analysis and some observations. J. Hydrol. 33, 133–146. Parker, G., 1978. Self formed rivers with stable banks and mobile bed. Part 1: The sand-silt river. J. Fluid Mech. 89, 109–125. Phillips, J.L., Ory, J., Talbot, A., 2000. Anadromous salmonid recovery in the Umatilla river basin, Oregon: a case study. J. Am. Water Resour. Assoc. 36, 1287–1308. Poff, N.L., Allan, J.D., Bain, M.B., et al., 1997. The natural ﬂow regime: a paradigm for river conservation and restoration. BioScience 47, 769–784. Puijalon, S., Bornette, G., 2004. Morphological variation of two taxonomically distant plant species along a natural ﬂow velocity gradient. New Phytol. 163, 651–660. Rao, A.R., Voeller, T., Delleur, J.W., Spacie, A., 1993. Estimation of instream ﬂow requirements for ﬁsh. J. Environ. Syst. 22, 381–396. Reiser, D.W., Wesche, T.A., Estes, C., 1989. Status of instream ﬂow legislation and practices in North America. Fisheries 24, 24–26. Rhoads, B.L., 1991. A continuously varying parameter model of downstream hydraulic geometry. Water Resour. Res. 27, 1865–1872. Rhoads, B.L., 1994. Fluvial geomorphology. Prog. Phys. Geogr. 18, 103–123. Ridenour, G.S., 2001. Interbasin consistency of Hydraulic geometry and its relationships with basin morphology and hydrology. Water Intl. 26, 569–577. Sagnes, P., Gaudin, P., Statzner, B., 1997. Shifts in morphometrics and their relation to hydrodynamic potential and habitat use during grayling ontogenesis. J. Fish Biol. 50, 846–858. Singh, K.P., McConkey Broeren, S., 1989. Hydraulic geometry of streams and stream habitat assessment. J. Water Resour. Plann. Manage. 115, 583–597. Statzner, B., Mu¨ller, R., 1989. Standard hemispheres as indicators of ﬂow characteristics in lotic benthos research. Freshwat. Biol. 21, 445–459. Statzner, B., Sagnes, P., Champagne, J.Y., Viboud, S., 2003. Contribution of benthic ﬁsh to the patch dynamics of gravel and sand transport in streams. Water Resour. Res. 39, 1–17. Stewardson, M.J., 2005. Hydraulic geometry of stream reaches. J. Hydrol. 306, 97–111. Stewardson, M.J., McMahon, T.A., 2002. A stochastic model of hydraulic variations within stream channels. Water Resour. Res. 38, 8.1–8.14.

Discussion by Rob Ferguson Your paper discusses statistical models for the distribution within a reach of point values of individual hydraulic variables (depth, velocity, shear stress). Is it possible to extend this approach and model the joint distribution of two variables, for example depth and velocity?

Hydraulic geometry of stream reaches and ecological implications

675

Nelson, R.L., Platts, W.S., Larsen, D.P., Jensen, S.E., 1992. Trout distribution and habitat in relation to geology and geomorphology in the North Fork Humboldt River drainage, northeastern Nevada. Trans. Am. Fish. Soc. 121, 405–426. Niemi, G.J., De Vore, P., Detenbeck, N., et al., 1990. Overview of case studies on recovery of aquatic systems from disturbance. Environ. Manage. 14, 571–587. Nikora, V.I., Aberle, J., Biggs, B.J.F., et al., 2003. Effects of ﬁsh size, time-to-fatigue and turbulence on swimming performance: a case study of Galaxias maculatus. J. Fish Biol. 63, 1365–1382. Orth, D.J., Maughan, O.E., 1982. Evaluation of the incremental methodology for recommending instream ﬂow for ﬁshes. Trans. Am. Fish. Soc. 111, 413–445. Osborne, L.L., Wiley, M.J., Larimore, R.W., 1988. Assessment of the water surface proﬁle model: accuracy of predicted instream ﬁsh habitat conditions in low-gradient, warmwater streams. Regul. Rivers Res. Manage. 2, 619–631. Park, C.C., 1977. World-wide variations in hydraulic geometry exponents of stream channels: an analysis and some observations. J. Hydrol. 33, 133–146. Parker, G., 1978. Self formed rivers with stable banks and mobile bed. Part 1: The sand-silt river. J. Fluid Mech. 89, 109–125. Phillips, J.L., Ory, J., Talbot, A., 2000. Anadromous salmonid recovery in the Umatilla river basin, Oregon: a case study. J. Am. Water Resour. Assoc. 36, 1287–1308. Poff, N.L., Allan, J.D., Bain, M.B., et al., 1997. The natural ﬂow regime: a paradigm for river conservation and restoration. BioScience 47, 769–784. Puijalon, S., Bornette, G., 2004. Morphological variation of two taxonomically distant plant species along a natural ﬂow velocity gradient. New Phytol. 163, 651–660. Rao, A.R., Voeller, T., Delleur, J.W., Spacie, A., 1993. Estimation of instream ﬂow requirements for ﬁsh. J. Environ. Syst. 22, 381–396. Reiser, D.W., Wesche, T.A., Estes, C., 1989. Status of instream ﬂow legislation and practices in North America. Fisheries 24, 24–26. Rhoads, B.L., 1991. A continuously varying parameter model of downstream hydraulic geometry. Water Resour. Res. 27, 1865–1872. Rhoads, B.L., 1994. Fluvial geomorphology. Prog. Phys. Geogr. 18, 103–123. Ridenour, G.S., 2001. Interbasin consistency of Hydraulic geometry and its relationships with basin morphology and hydrology. Water Intl. 26, 569–577. Sagnes, P., Gaudin, P., Statzner, B., 1997. Shifts in morphometrics and their relation to hydrodynamic potential and habitat use during grayling ontogenesis. J. Fish Biol. 50, 846–858. Singh, K.P., McConkey Broeren, S., 1989. Hydraulic geometry of streams and stream habitat assessment. J. Water Resour. Plann. Manage. 115, 583–597. Statzner, B., Mu¨ller, R., 1989. Standard hemispheres as indicators of ﬂow characteristics in lotic benthos research. Freshwat. Biol. 21, 445–459. Statzner, B., Sagnes, P., Champagne, J.Y., Viboud, S., 2003. Contribution of benthic ﬁsh to the patch dynamics of gravel and sand transport in streams. Water Resour. Res. 39, 1–17. Stewardson, M.J., 2005. Hydraulic geometry of stream reaches. J. Hydrol. 306, 97–111. Stewardson, M.J., McMahon, T.A., 2002. A stochastic model of hydraulic variations within stream channels. Water Resour. Res. 38, 8.1–8.14.

Discussion by Rob Ferguson Your paper discusses statistical models for the distribution within a reach of point values of individual hydraulic variables (depth, velocity, shear stress). Is it possible to extend this approach and model the joint distribution of two variables, for example depth and velocity?

676

N. Lamouroux

Reply by the author Stewardson and McMahon (2002) proposed a model of the joint depth and velocity probability distribution in stream channels. I do not know of other published work on the subject. However, I am aware of ongoing research for extending the statistical approach to multivariate distributions and the spatial arrangement of hydraulic variables. Such efforts certainly contribute to increase the range of applications of statistical hydraulic models. One difﬁculty may be that the correlation between, for example point depth and point velocity can be highly variable across stream reaches and may depend strongly on local features (e.g., a large boulder, a vegetated bank).

Gravel-Bed Rivers VI: From Process Understanding to River Restoration H. Habersack, H. Pie´gay, M. Rinaldi, Editors r 2008 Elsevier B.V. All rights reserved.

677

26 Gravel bars: a key habitat of gravel-bed rivers for vegetation David Gilvear, Robert Francis, Nigel Willby and Angela Gurnell

Abstract This papers examines the importance of gravel bars in terms of a substrate for recruitment, colonisation and development of ground ﬂora and woody vegetation via two European case studies. Experimental work on the River Tagliamento in Italy is used to explore the role of substrate particle size and elevation on recruitment and growth of seedlings and cuttings (Populus nigra L. and Salix elaeagnos Scop.). Meanwhile on the River Tummel in Scotland, the pattern of vegetation communities are related to bar morphology and sedimentology via ﬁeld survey. Both studies reveal the critical importance of bar morphology and substrate particle size, via their control on inundation frequency, substrate stability and moisture availability, in terms of vegetation development on gravel bars from the initial colonisation stage to vegetation communities present after more than decade.

1.

Introduction

Gravel bars and partially vegetated islands are key components of gravel-bed rivers (Poff and Ward, 1990; Edwards et al., 1999; Tockner et al., 2003) and are naturally highly dynamic and morphologically complex systems. Natural bar features often contain highly heterogeneous sediments and relatively large topographical ranges (Lewin, 1996; Van Coller et al., 2000) and during ﬂoods can be exposed to simultaneous sediment erosion and deposition and to hydrochorous propagule deposition. As such the ﬂuvial bar environment has a profound inﬂuence on the establishment and development of riparian plant species, inﬂuencing as it does patterns of vegetation destruction and burial, and regulating availability of moisture and nutrients via the porosity and permeability, and depth to the water table within the substrate. Gravel bars therefore represent important instream habitat at the aquatic terrestrial interface. In having an ecotonal position between aquatic and terrestrial environments, bars have the potential to support a varied and rich biota. In the UK, for example, the importance of exposed riverine sediments for ground beetle diversity is E-mail address: [email protected] (D. Gilvear) ISSN: 0928-2025

DOI: 10.1016/S0928-2025(07)11154-8

678

D. Gilvear et al.

now recognised (Eyre et al., 2001; Sadlar et al., 2004). One interpretation of this linkage is that vegetation dynamics on islands and bars is effectively a barometer of functional integrity and more general ﬂoodplain ecosystem health. However, river modiﬁcation due to navigation, ﬂood control engineering and the general containment of natural geomorphological processes now means that islands represent an increasingly endangered attribute of ﬂoodplain corridors. Consequently there is a need to assess the functional significance of these landforms at a ﬂoodplain level. An understanding of the interaction between gravel bar surfaces and vegetation dynamics is of scientiﬁc interest and essential to the successful reinstatement of more natural riparian communities on heavily modiﬁed river corridors, and the restoration of associated biodiversity and function. This is particularly important for many rivers in the Alps and other mountainous regions of western Europe where gravel bar development and channel migration has been extensively constrained by extensive bank protection, river training, gravel trapping and removal and ﬂood defence structures. Simons and Simons (1987) view river bars as large bedforms resulting from sediment deposition. Their classiﬁcation and evolution has been the subject of many investigations by geomorphologists. For example, on meandering rivers, depositional features on the inside of bends and attached to the ﬂoodplain have been identiﬁed as point bars. Their evolution has been mapped by using the pattern, often identiﬁed via the distribution of vegetation, of ﬂoodplain ridge and swale left by migrating point bars (Nanson and Croke, 1992). Mid-channel bars are more typical of braided and wandering planforms and tend to be unvegetated due to rapid evolution and frequent bed instability. Brice (1964) deﬁned mid channel bars as being unvegetated and submerged at bankfull whereas islands are vegetated and emergent at bankfull stage. In reality this distinction is blurred because the low-water season annual plant species can establish on otherwise active bars and mid-channel bars are often a response to gravel accumulation in the lee of ﬂoodborne trees stranded during falling stage (Gurnell et al., 2001). A simple model of bar development was proposed by Jaegii (1987). In the ﬁrst phase, sediment is deposited until a limiting height is achieved. In phase two, material is deposited in the lee of the initial phase as a tail. However, bars should not be thought of as single morphological/sedimentological entities. They often exist as the result of a complex erosional and depositional chronology linked to the nature of the ﬂood series following bar initiation. Discrete morphological units such as bar heads, bar lobes, avalanche faces, bar tails, cross-bar channels and sloughs can be identiﬁed. Consequently many bars show considerable internal topographic, sedimentological and chronological variability. The degree of vegetation development on bars is likely to be related to the amount of time the surface has been exposed above the seasonal low-water mark, the depth to the water table, the physical character of these sediments and their stability and the types of vegetation available for colonisation. Depending on these factors, newly formed bars are progressively vegetated as they accrete vertically and laterally and it thus becomes difﬁcult to deﬁne where a point bar becomes part of the ﬂoodplain and an exposed mid-channel bar becomes a wellvegetated island. The separation is in effect artiﬁcial but the fact that bar development is often the ﬁrst stage in island creation and ﬂoodplain evolution is important both geomorphologically and ecologically.

Gravel bars

679

Despite numerous botanical studies of ﬂoodplain and riparian landscapes (e.g., Hupp and Osterkamp, 1985; Kalliola and Puhakka, 1988; Prach, 1994; Hupp and Osterkamp, 1996; Tabbacchi et al., 1996; Girel and Manneville, 1998; Abernethy and Willby, 1999; Gilvear et al., 2000; Johnson, 2000; Gurnell et al., 2001) there has been little explicit focus on bars, particularly on gravel-bed rivers. A notable exception is the work of Bendix and Hupp (2000) who worked on the gravel bars of the Missouri River and established that there was a critical linkage between ﬂoods, propagule transport, bar formation and cottonwood forest development and riparian species richness. Similarly, Dykaar and Wigington (2000) working on the Willamette River in Oregon traced patterns of cottonwood ﬂoodplain forest back to the evolution of underlying bar forms. In Europe, the interaction between gravel bar geomorphology and vegetation development has been restricted to relatively few studies on a restricted number of rivers (e.g., Marston et al., 1995; Steiger and Gurnell, 2003; Gilvear and Willby, 2006) most notably the River Tagliamento (e.g., Kollmann et al., 1999; Karrenberg et al., 2003; Tockner et al., 2003; Francis et al., 2004; Francis et al., 2006). Instream and riparian plant species diversity depends on one hand on the inﬂuence of scale-dependent processes including disturbance (i.e., ﬂooding, erosion and sedimentation), physical stresses (i.e., drought or waterlogging) and on the other biotic interactions (i.e., interspecific competition for light or nutrients) and population dynamics, seed dispersal and regeneration. The resulting landforms inﬂuence patterns of vegetation development through a multitude of controls including elevation and hence frequency, depth, and duration of ﬂooding, various soil properties and in particular grain size and moisture content (Barnes, 1978; Prach, 1994; Robertson and Augspurger, 1999; Scott et al., 1999; Shin and Nakamura, 2005). These studies suggest that elevation may be the best explanatory variable for herbaceous species distribution. Moreover, by creating ﬂuvial surfaces of differing age discrete plant communities with similar species composition but different age classes exist in close juxtaposition. Thus it is likely that physical habitat heterogeneity and ﬂuvial disturbance are the over-riding local controls on gravel-bar plant diversity. Their role however, is conditional upon the importance of the composition of the regional species pool and the attendant effects of population fragmentation and site connectivity on opportunities for dispersal and colonisation from lateral and upstream sources (Johnson, 2002). The importance of understanding the various abiotic controls on ﬂoodplain vegetation and ability to model the role of channel change in controlling river and riparian vegetation dynamics has been highlighted (Ward et al., 2001; Richards et al., 2003). Coupled geomorphological–ecological models sensitive to the effects of evolving channel morphology and levels of ﬂuvially induced instability, are needed for river restoration and management purposes. This paper will explore the role of gravel bar elevation and grain size and the linked processes of ﬂood disturbance and moisture deficit together with plant phenology on plant colonisation and development in two contrasting European gravel-bed river environments; namely the River Tagliamento draining the alpine region of North East Italy (Fig. 26.1) and the Tummel draining the central highlands of Scotland (Parsons and Gilvear, 2002; Gilvear and Willby, 2006). The ﬁrst study presented will concentrate on the recruitment and survival of woody species while the

680

D. Gilvear et al.

Figure 26.1. The River Tagliamento, NE Italy. The river catchment is delineated by the dashed line, and the location of the study reach for the ﬁeld experiment is highlighted.

second will examine in more detail herbaceous vegetation distribution and factors inﬂuencing seral stages. On the Tagliamento two key riparian tree species (Populus nigra L. and Salix elaeagnos Scop.) were monitored over two growing seasons, and investigations focused not only on how the bar environment affected different species, but also how differences in propagule form (seeds vs. vegetative fragments/ cuttings) inﬂuenced responses. On the River Tummel, the role of elevation and particle size across a single complex gravel bar formation in determining the pattern and composition of vegetation is studied. Here the role of ﬂood dynamics in creating

Gravel bars

681

ﬂuvial surfaces with different geomorphic and hydrological characteristics and patterns of vegetation development is explored. By combining the ﬁndings of these two contrasting case studies, each with their own objectives, a picture of the role of bar topography and grain size on the establishment and development of vegetation on gravel bar surfaces will emerge.

2.

The River Tagliamento case study

The braided River Tagliamento was the location for a ﬁeld experiment into the inﬂuence of sediment calibre and topography on the establishment of two important riparian tree species (P. nigra L. and S. elaeagnos Scop.). The river runs approximately 170 km from its source in the Italian Alps to the Adriatic Sea, and has a catchment of 2850 km2. For the most part, the river maintains natural functional and morphological dynamics (e.g., Tockner et al., 2003). Two relevant aspects of the river are: (1) the amount (approximately 90%) of active zone sediments exposed for the majority of the year, and (2) the abundance of large bars (100–1000 m long). High ﬂows generally occur within spring (resulting from snowmelt) and autumn (due to periods of intense precipitation), with low ﬂows most common between December and February, and during August (Gurnell et al., 2000). The ﬁeld investigations reported here took place in 2002–2003 within the active zone of a prealpine reach 80 km from the source and at an altitude of approximately 140 m above sea level. 2.1.

Methodology

The experiment involved the planting of cuttings of P. nigra and S. elaeagnos in three stages designed to incorporate their deposition at varying points within the growing season. Stage one planting took place in April, during which six plots of 100 cuttings (50 P. nigra and 50 S. elaeagnos in each plot) were set up (Fig. 26.2; Table 26.1) in locations selected to broadly differ in sedimentary and elevational characteristics (and nominally classiﬁed as High Fine, Low Fine, High Mixed, Low Mixed, High Coarse and Low Coarse). Stage two planting took place in July, when a further three plots (at high elevations) were set up adjacent to the existing High Stage one plots. This stage also involved the selection of plots containing 1–2 year old seedlings of the two subject species in comparable locations to the Stage one cutting plots, so that seedling growth could be compared to that observed from cuttings. Stage three cutting planting took place in September, when a further three High plots were set up next to their Stages one and two counterparts. Stages two and three involved planting only at high plots due to losses of cuttings observed at Stage one low plots before Stage two planting. For all stages, relative elevation of the plots was determined and corrected for ﬂoodplain slope. Surface sediment calibre was assessed using a photosieving technique (see Petts et al., 2000), while subsurface grain-size distribution was determined using both dry-sieving and laser grain sizing. Subsurface samples were taken from a depth of 25 cm as this was considered the key depth for determining moisture availability around young roots and root primordia during the

D. Gilvear et al.

682

Figure 26.2. An experimental plot of cuttings on the Tagliamento river.

initial phases of establishment, which is affected by grain size (Moreton et al., 2002). Percentage organic matter by weight on samples taken to determine sediment grain size was determined by loss on ignition. 2.1.1.

Cutting selection and planting

For all stages, cuttings were taken from several individuals of both species situated on islands or within the ﬂoodplain woodland close to the plots. Each cutting was selected from old-growth branches and cut to a length of 40 cm, with a diameter of X5 mm for at least 30 cm of the length. This was a standardized length and diameter to limit variations in size, which may affect survival and growth of cuttings (Dickmann et al., 1980; Burgess et al., 1990). As several individuals were used, the cuttings were mixed to avoid any genetic or sexual bias. Cuttings were planted in May or July with a 0.5 m spacing in each plot (ten rows of ten cuttings in each stage), and species were planted in alternate rows. Each cutting was planted, with minimal disturbance, at an angle of approximately 451 to the ground surface, pointing downstream. Less than 10 cm of each cutting was left protruding from the sediment after planting. Each plot was then watered for 3 days to simulate conditions during the falling limb of a ﬂood on the river, when plant fragments would ordinarily be ﬂuvially deposited. 2.1.2.

Cutting and seedling growth measurements

Survival and growth measurements were taken during 12–16 July 2002, 25–27 September 2002, 14–17 April 2003 and 18–20 August 2003. Cuttings were considered

Gravel bars

683

Table 26.1. Summary of the plots set up in a ﬁeld experiment along the River Tagliamento, including dates of planting (or marking in the case of seedling plots) and dates when survival and growth measurements were taken. Plot

Date of planting/marking

Dates of measurement

Stage one High ﬁne

27/04/2002–03/05/2002

Low ﬁne

27/04/2002–03/05/2002

High mixed

27/04/2002–03/05/2002

Low mixed

27/04/2002–03/05/2002

High coarse

27/04/2002–03/05/2002

Low coarse

27/04/2002–03/05/2002

12–16/07/2002, 25–27/09/ 2002, 14–17/04/2003, 18–20/08/2003 12–16/07/2002, 25–27/09/ 2002, 14–17/04/2003, 18–20/08/2003 12–16/07/2002, 25–27/09/ 2002, 14–17/04/2003, 18–20/08/2003 12–16/07/2002, 25–27/09/ 2002, 14–17/04/2003, 18–20/08/2003 12–16/07/2002, 25–27/09/ 2002, 14–17/04/2003, 18–20/08/2003 12–16/07/2002, 25–27/09/ 2002, 14–17/04/2003, 18–20/08/2003

Stage two High ﬁne

17/07/2002–19/07/2002

High mixed

17/07/2002–19/07/2002

High coarse

17/07/2002–19/07/2002

Stage three High ﬁne

07/09/2002–09/09/2002

High mixed

07/09/2002–09/09/2002

High coarse

07/09/2002–09/09/2002

Seedlings High ﬁne Low ﬁne High mixed Low mixed High coarse Low coarse

20/07/2002–25/07/2002 20/07/2002–25/07/2002 20/07/2002–25/07/2002 20/07/2002–25/07/2002 20/07/2002–25/07/2002 20/07/2002–25/07/2002

Comments

Eroded completely on 19th November 2002

Eroded completely on 19th November 2002 Eroded completely on 19th November 2002

25–27/09/2002, 14–17/04/ 2003, 18–20/08/2003 25–27/09/2002, 14–17/04/ 2003, 18–20/08/2003 25–27/09/2002, 14–17/04/ 2003, 18–20/08/2003

Eroded completely on 19th November 2002

14–17/04/2003, 18–20/08/ 2003 14–17/04/2003, 18–20/08/ 2003 14–17/04/2003, 18–20/08/ 2003

Eroded completely on 19th November 2002

20/07/2002–25/07/2002 20/07/2002–25/07/2002 20/07/2002–25/07/2002 20/07/2002–25/07/2002 20/07/2002–25/07/2002 20/07/2002–25/07/2002

alive if they displayed any living stems. Growth measurements included: (1) number of stems, (2) total combined length of all stems, (3) number of leaves and (4) average leaf length (based on the mean of a sample of ten leaves). Seedling measurements consisted of: (1) seedling age, (2) number of stems, (3) total stem length and (4) number of leaves. P. nigra is monopodial with an apical meristem, and leaves a distinct winter bud scar – consequently, seedlings of this species were aged by counting these scars. S. elaeagnos is sympodial (no apical meristem) and so this species was aged by counting leaf bud scars. When estimating ages of seedlings in days, growth was assumed to commence on 1 May and, for 2-year-old seedlings, to have ceased between 1 October and 30 April inclusive.

D. Gilvear et al.

684

Table 26.2. Substrate characteristics for cutting and seedlings plots set up along the River Tagliamento. Depth Cuttings high ﬁne Surface Subsurface (25 cm depth) Seedlings high ﬁne Surface Subsurface (25 cm depth) Cuttings low ﬁne Surface Subsurface (25 cm depth) Seedlings low ﬁne Surface Subsurface (25 cm depth) Cuttings high mixed Surface Subsurface (25 cm depth) Seedlings high mixed Surface Subsurface (25 cm depth) Cuttings low mixed Surface Subsurface (25 cm depth) Seedlings low mixed Surface Subsurface (25 cm depth) Cuttings high coarse Surface Subsurface (25 cm depth) Seedlings high coarse Surface Subsurface (25 cm depth) Cuttings low coarse Surface Subsurface (25 cm depth) Seedlings low coarse Surface Subsurface (25 cm depth)

D5

D50

D95

Gravel (%)

Sand (%)

Organic (%)

Elevation

(43) 2.9

(43) 1.5

(43) 0.8

0 12.6

100 87.4

– 0.7

96 –

(43) 2.55

(43) o1

(43) o1

0 73.1

100 26.6

– 0.8

190 –

(43) 3

(43) 1.6

(43) 0.1

0 0.8

100 99.2

– 1.1

51 –

(43) 44

(43) 2.55

(43) 0.65

0 0.4

100 99.6

– 2.0

38 –

(43) 2.5

3.3 o1

5.3 o1

– 68.3

– 31.7

– 0.6

191 –

(43) 2.4

5.2 o1

7.1 o1

76.5 59.6

23.5 40.3

– 0.8

185 –

(43) 2.8

(43) o1

5.6 o1

– 74.8

– 25.2

– 1.4

54 –

(43) 2.85

4.8 o1

6.95 o1

80 82.4

20 17.6

– 0.1

58 –

(43) 1.6

4.8 o1

6.5 o1

– 86.2

– 13.8

– 0.4

125 –

(43) 2.1

6.05 o1

7.5 o1

87.8 66.8

12.3 33.2

– 0.9

116 –

(43) 1.9

5 o1

6.9 o1

– 80.5

– 19.5

– 0.9

83 –

(43) 2.45

4.8 o1

8.4 o1

84.5 78.7

15.5 21.3

– 0.8

72 –

Note: Elevation in centimetres is expressed relative to stage at low ﬂow in a nearby channel during summer 2003. Particle size measurements are in Phi units.

Gravel bars 2.2. 2.2.1.

685

The results The physical environment

Table 26.2 summarises the physical environmental characteristics of the cutting and seedling plots. Given these ﬁndings it was concluded that any significant differences in plant survival and growth could be considered reliable indicators of the effects of sediment calibre and relative elevation. 2.2.2.

Cutting survival

In Stage one, survivorship was higher for S. elaeagnos over P. nigra for all measurement periods (Table 26.3), and following Mann–Whitney tests with Bonferroni correction was found to be significantly (Po 0.05) higher in the High Fine, High Mixed, Low Mixed and Low Coarse plots after being planted for 75 days (July 2002) and 150 days (September 2002). Survivorship in subsequent measurement periods could not be compared statistically due to the low number of survivors. The High Fine plot produced no survivors of P. nigra. Within Stage two, initial survival after 75 days (September 2002) was high for both species within all three plots (High Fine, High Mixed and High Coarse), although subsequent survivorship of both species was very low, with few cuttings surviving the winter period (Table 26.3). Stage three cuttings, planted in September 2002 and therefore at the end of the growing season, produced several survivors of both species for the High Fine and High Mixed plots (the High Coarse plot having been lost during the November 2002 ﬂood), and the number of survivors increased substantially by August 2003 (330 days after planting). The increase in survivorship between April 2003 and August 2003 probably represent sprouts from buried or late-sprouting buds that were not apparent at the time of the April 2003 survey; although some cuttings in the experiment were observed to sprout several months after deposition (such as in the Stage one Low Coarse plot for P. nigra) (Table 26.3). By August 2003, September (Stage three) planting had produced many more P. nigra survivors than S. elaeagnos (almost double; Table 26.3). It can therefore be seen that time of planting/deposition can affect the survivorship of each species from vegetative fragments. Results suggest that S. elaeagnos cuttings can survive relatively easily at any time of deposition, whereas P. nigra displays much greater survivorship when deposited near the end of the growing season, probably due to the higher carbohydrate and nitrogen resources contained within the fragment, which are built up earlier in the growing season. For both species, deposition within the middle of the growing season produced the highest initial survivorship, but the lowest overall survivorship. 2.2.3.

Cutting growth

To evaluate the inﬂuence of relative elevation on the growth of each species, for the July and September 2002 measurements the growth data from the High and Low plots were pooled into single High and Low datasets respectively, and then Mann–Whitney tests performed with Bonferroni correction (Table 26.4A). This was then repeated for sediment calibre, pooling data for Fine, Mixed and Coarse sites as

686

Table 26.3.

Percentage survivorship of cuttings at various times of measurement for all stages of the experiment.

Time of measurement (days since planting)

HF

HM

HC

S. elaeagnos P. nigra S. elaeagnos P. nigra Stage one (May 2002 planting) July 2002 (75) 32 0 September 2002 28 0 (150) April 2003 (350) 18 0 August 2003 (475) 12 0 Stage two (July 2002 planting) September 2002 42 60 (75) April 2003 (275) 4 0 August 2003 (400) 4 0 Stage three (September 2002 planting) April 2003 (200) 32 24 August 2003 (330) 42 54

LF

LM

LC

Total (%)

S. elaeagnos

P. nigra

S. elaeagnos

P. nigra S. elaeagnos

P. nigra S. elaeagnos P. nigra S. elaeagnos

P. nigra

13 11

60 58

12 12

12 12

18 18

6 *

14 *

74 72

16 16

56 56

18 20

40 37.7

12 38

20 12

* *

* *

* *

* *

* *

* *

20 20

6 16

8.3 8.3

4.7 4.7

50

50

68

50

–

–

–

–

–

–

53.3

53.3

34 12

2 6

* *

* *

– –

– –

– –

– –

– –

– –

12.7 5.3

0.7 2

14 24

34 76

* *

* *

– –

– –

– –

– –

– –

– –

15.3 22

19.3 43.3

*Indicates a plot wherein all cuttings were destroyed by ﬂooding. – indicates that plots of this classiﬁcation were not set up in a given Stage. H and L in experiments refers to High and Low elevation and F, M and C to ﬁne medium and coarse particle size.

D. Gilvear et al.

Gravel bars

687

Table 26.4. Mann–Whitney analyses comparing variations in growth parameters of surviving stage one Salix elaeagnos and Populus nigra cuttings in July and September 2002, according to (A) relative elevation and (B) sediment calibre. Growth parameter

Salix elaeagnos

Populus nigra P (adj for ties)

High vs. low July 2002

High vs. low September 2002

High vs. low July 2002

High vs. low September 2002

0.248 0.105 0.039 (High4Low) 0.548

0.459 0.154 0.181 0.000 (low4high)

0.007 0.022 0.001 0.043

0.000 (low4high) 0.058 0.147 0.002 (low4high)

Significant differences between sediment calibres July 2002

Significant differences Significant differences between sediment between sediment calibres July 2002 calibres September 2002

None Fine4Mixed Mixed4Coarse None Coarse4Fine

Coarse4Mixed None

None None

None None

None None

None Fine, Coarse4Mixed

Coarse4Mixed None

A Total stem length Number of stems Number of leaves Average length of leaves

(low4high) (Low4High) (low4high) (low4high)

B

Total stem length Number of stems Number of leaves Average length of leaves

Significant differences between sediment calibres September 2002

Note: Italicised text in (A) indicates a significant difference at the 0.05 level. Emboldened text in (A) indicates a significant difference at the 0.01 level. Emboldened text in (B) indicates a significant difference between sediment calibres at the 0.05 level, with Bonferroni correction.

appropriate (Table 26.4B). For the August 2003 measurements pooling was no longer an option, and sites were compared individually using the same statistical tests (Table 26.4A). Variations in relation to relative elevation were most marked for P. nigra, where the majority of growth parameters were significantly larger in the lower plots. By August 2003 the Low Coarse site had the highest growth rates for all growth parameters for both species. The number of statistically significant differences obtained for growth parameters at varying elevations was higher over both periods of measurement for P. nigra cuttings than the number obtained for S. elaeagnos cuttings. The opposite was true for significant differences obtained due to variations in sediment calibre, with many more being found for S. elaeagnos cuttings than P. nigra. In Stage two, September 2002 measurements demonstrated that for S. elaeagnos, growth rates were highest in the High Fine plot for total stem length and average leaf length, while the High Coarse plot had the greatest numbers of stems and leaves. For P. nigra, the High Mixed plot maintained the highest values for all growth parameters. August 2003 measurements found a reversal of the September 2002 trends for S. elaeagnos. The August 2003 measurements are based on very few surviving cuttings, and so contrasts from this period may not be reliable. Furthermore, the High Coarse plot had been destroyed by this point. Only one Stage two plot contained any surviving P. nigra by August 2003, and comparisons were not

D. Gilvear et al.

688

Table 26.5. Mann–Whitney analyses comparing variations in growth parameters of surviving (A) stage one and (B) stage three Salix elaeagnos and Populus nigra cuttings within remaining plots during August 2003. Salix elaeagnos

Populus nigra

Total stem length Number of stems Number of leaves Average length of leaves

LC4HF, HM LC4HM LC4HM LC4HF, HM

LC4HM LC4HM LC4HM None

Total stem length Number of stems Number of leaves Average length of leaves

None HF4HM None HM4HF

HM4HF None None HM4HF

Growth parameter A

B

Note: Emboldened text indicates a significant difference between sediment calibres at the 0.05 level, with Bonferroni correction. H and L in experiments refers to high and low elevation and F, M and C to ﬁne medium and coarse particle size.

possible. Indeed, statistical comparisons from Stage two were only feasible for the September 2002 measurements for S. elaeagnos, where cuttings within the High Coarse plot displayed significantly (po0.05) greater numbers of stems and leaves than those in the High Mixed plot, following Mann–Whitney analysis with Bonferroni correction. For Stage three, measured only in August 2003, the High Coarse plot had been destroyed and so comparisons could only be made between the High Fine and High Mixed plots (Table 26.5B). For S. elaeagnos, the High Mixed plots contained cuttings experiencing the greatest growth for all parameters except number of stems, while for P. nigra the cuttings in the High Mixed plot displayed the highest growth rates for all parameters without exception. Statistical comparisons (Table 26.5B) found that for S. elaeagnos, the High Fine plot had a significantly (po0.05) greater number of stems than the High Mixed plot, while the High Mixed plot had a significantly greater average leaf length than the High Fine plot. For P. nigra, the High Mixed plot contained significantly (po0.05) greater total stem, longest stem and average leaf lengths than in the High Fine plot. Overall total stem growth (day1) of cuttings during each Stage is given in Table 26.6. A trend for decreasing daily growth over time is apparent in the cuttings, with growth occurring at a much faster rate for both species in the ﬁrst growing season after deposition. Growth rates were greater in S. elaeagnos over P. nigra at all times of measurement. 2.2.4.

Seedling growth

No S. elaeagnos seedlings were found within the High Fine and High Mixed plots. For S. elaeagnos classiﬁed as 1 year, the highest growth rates were in the High Coarse plot, while for 2-year seedlings, Low Coarse performed best. For P. nigra

Gravel bars

689

Table 26.6. Average total stem length increment (mm day1) for S. elaeagnos and P. nigra cuttings and seedlings in all stages. Time of measurement (days of growth)

S. elaeagnos

P. nigra

Stage one cuttings July 2002 (75) September 2002 (150) August 2003 (265)

4.0 3.2 0.8

1.9 1.5 0.5

Stage two cuttings September 2002 (75) August 2003 (190)

1.6 0.6

0.8 0.5

Stage three cuttings August (115)

3.8

3.0

Seedlings Age 1 (75) Age 2 (230)

1.2 1.4

0.8 0.6

seedlings, growth rates were highest in the Low Coarse plot for both 1- and 2-year seedlings. A positive relationship was observed between growth rates and increasing sediment coarseness for many variables of both S. elaeagnos and P. nigra, with the exception of the Low plots for P. nigra, where the relationships were generally negative. Data were grouped according to elevation and sediment calibre and analysed using Mann–Whitney tests with Bonferroni correction to determine if there were any significant differences between growth rates of seedlings according to these variables, for each age class (1 or 2 years) (Table 26.7). Although S. elaeagnos seedlings were not present in the High Fine and the High Mixed plots, those found on the High Coarse plot that were aged 2 years performed significantly better than those in the grouped Low plots for all three measured growth parameters. The grouped P. nigra data also displayed a trend for significantly higher growth rates in the High plots for year 2 seedlings (Table 26.7). In relation to sediment calibre, S. elaeagnos seedlings displayed significantly higher growth rates in the Coarse plots than in the other plots, while second-year P. nigra seedlings performed better in the Fine plots. Average daily growth rates for total stem length of seedlings of both species and ages are given in Table 26.6. Daily increment was not substantially different for seedlings of either age, suggesting a more linear rate of the growth than that found in cuttings (Table 26.6). S. elaeagnos seedlings out-performed P. nigra seedlings.

3.

The River Tummel case study

The River Tummel in its lower reaches is typically 60 m wide and ﬂows within a wandering gravel-bed channel. Here the River Tummel has a mean discharge of 70 m3 s1. The reach lies some 2 km from the conﬂuence with the River Tay. In this

D. Gilvear et al.

690

Table 26.7. (A) Mann–Whitney analyses comparing variations in growth parameters of ﬁrst year and second year Salix elaeagnos and Populus nigra seedlings, surveyed in July 2002, according to (A) relative elevation and (b) sediment calibre. Growth parameter

Salix elaeagnos

Populus nigra

P (adj for ties)

P (adj for ties)

High versus low age 1

High versus low age 2

High versus low age 1

High versus low age 2

0.179 0.332

0.000 (high4low) 0.000 (high4low)

0.943 0.455

0.001 (high4low) 0.001 (high4low)

Significant differences between sediment calibres age 1

Significant differences between sediment calibres age 2

Significant differences between sediment calibres age 1

Significant differences between sediment calibres age 2

Mixed, Coarse4Fine Coarse4Mixed Mixed, Coarse4Fine

Coarse4Fine, Mixed

None

Fine4Mixed

Coarse4Fine, Mixed

None

Fine4Mixed

A Total stem length Number of leaves B

Total stem length Number of leaves

Note: Italicised text in (A) indicates a significant difference at the 0.05 level. Emboldened text indicates a significant difference at the 0.01 level. Emboldened text in (B) indicates a significant difference between sediment calibres at the 0.05 level, with Bonferroni correction.

reach, a gravel bar complex has evolved (Fig. 26.3) over the last 25 years with discrete geomorphic units, each in part being attributable to the geomorphic impact of a series of large ﬂoods that has occurred in recent years. A large percentage of the now partially colonised surfaces were bare gravel in 1993. One area in 1990 and for a number of years preceding this date was well vegetated but most of the island was over-ridden by gravels during ﬂoods in the late 1980s and early 1990s. The entire gravel bar complex is less thlan 2 ha in size. Its maximum elevation range is 2.8 m. The bar head ranges from between a few centimetres and 2.5 m above summer low ﬂow. The bar tails lie at an elevation of between 1.5 and 2.0 m below the maximum elevation of the bar head and up to 0.50 to 0.65 m above summer low ﬂow. Mean sediment size for the armour layer of gravel samples varied between 21 and 64 mm (Table 26.8). Eighty-fourth percentile values varied between 15 and 118 mm. Some gravel areas were totally draped with sand and recorded as such. Moisture levels in the sediments ranged from less than 5% to 35%. Organic content varied between less than 1% and 15% but values were generally below 3%.

3.1. 3.1.1.

Methodology Field studies

Twelve transects were set up across the gravel bar complex to encompass the lateral and longitudinal morphological, sedimentological and ﬂoristic variability. Sampling points were roughly equally spaced 15 m apart; in total 66 sampling points were established. All sample points were surveyed with x, y, z co-ordinates and corrected for water slope. The vegetation was surveyed and recorded 2 2 m quadrats. All

Gravel bars

691

Figure 26.3. The River Tummel gravel bar morphology and spatial arrangement of vegetation clusters identiﬁed. Cluster 1 – mid elevation dry coarse gravels supporting sparse pioneer vegetation; cluster 2 – high-elevation coarse gravels supporting mix of herbs, bryophytes and grasses with Saronthamnus scorparius and U. europaeus with occasional Acer pseudoplatanus saplings; cluster 3 – mid elevation dry moderate grain size gravels supporting mixed herb-rich mesotrophic grassland; cluster 4 – ﬁner damp gravels supporting mixed herb/shaded neutral grassland; cluster 5 – wet organic low-lying sediments and sand splays supporting Salix spp. Bushes.

species present were recorded and their abundance assessed as percent cover. All sampling was undertaken during two dry days in late July 2002. The grain size of the surface layer of the exposed bar sediments was quantiﬁed by randomly sampling 100 grains within the 2 2 m quadrat, grains being measured using a graduated ‘‘pebble plate’’. The surface layer was also cleared over a 0.3 0.3 m area and a 2–4 kg sample of the underlying substrate collected. The sample was passed through a 32 mm sieve in the ﬁeld to remove the coarsest sediment fraction. The sample was

D. Gilvear et al.

692

Table 26.8. Environmental and botanical characteristics of each cluster identiﬁed on the River Tummel gravel bar. Cluster

Elevation (m) above low stage

Sediment size (mm)

Sediment sorting

Moisture content (%)

Organic content (%)

Armouring size (mm)

Armouring sorting

1 2 3 4 5

0.36 0.73 0.31 0.32 0.22

1.6 1.7 1.2 0.3 0.6

1.4 1.4 1.2 0.1 0.7

2.1 4.1 3.2 14.5 34.6

1.1 3.1 1.3 2.0 10.0

55.4 51.4 27.8 11.5 0.5

27.4 25.9 12.6 5.2 0.7

Cluster

Vegetation cover (%)

Sample richness

Number of samples

Standardised species pool

Error

Beta index

Unique species

1 2 3 4 5

20.3 45.5 30.0 49.6 50.0

10.4 19.1 14.7 18.5 18.0

11 16 23 12 4

55.6 76.5 80.7 94.7 74.0

1.09 2.14 3.50 1.39 N/A

0.53 0.40 0.55 0.51 0.40

2.9 19.5 11.8 9.8 7.0

Note: All values are means based on the number of samples per cluster. Sorting coefﬁcients were calculated using D16/D84. Sample richness refers to alpha diversity (i.e. the mean number of species per 2 2 m quadrat). The species pool for each cluster is based on a standard sample size of 10 calculated using rarefaction (Coleman estimate) or extrapolation (cluster 5 only). The beta index, a measure of turnover between samples, ¼ species pool/(sample richness 10). The number of unique species is based on the number of species observed only in one cluster scaled relative to the size of the estimated species pool.

then sealed for laboratory determination of grain size distribution, and organic and moisture content. 3.1.2.

Laboratory analysis

Samples of the sub-surface gravels were analysed for three physical properties. Soil moisture content was quantiﬁed by drying at 110 1C for 2 h. Organic matter was determined by loss on ignition. Grain size analysis was by dry sieving down to 63 mm. 3.1.3.

Data analysis

The environmental data was classiﬁed into ﬁve groups using cluster analysis (based on average linkage) within MINITAB v 12. Elevation, mean grain size and sorting of the armoured layer, soil moisture content, percentage organic matter, and sediment grain size formed the input variables. The mean sample species richness, and the frequency and abundance of plant species in each of the ﬁve clusters were then calculated. Species unique to each cluster were also identiﬁed. To estimate the species pool associated with each cluster based on a constant sampling effort extrapolation and rarefaction, as appropriate to the number of samples per cluster, were undertaken using Estimate S v 5.01 (Colwell, 1997). This enabled a comparison of species turnover (beta diversity) between samples within each of the different clusters and an

Gravel bars

693

assessment of the relative contribution of each cluster to the species pool associated with the mosaic of habitats on the site as a whole. Species–environmental relationships were analysed using Canonical Correspondence Analysis (CCA) with CANOCO v 4.5. The inﬂuence of variables was tested using manual forward selection supported by Monte Carlo random permutation tests (999 runs) of the significance of each additional variable.

3.2. 3.2.1.

Results Vegetation and plant diversity

The environmental characteristics of each of the ﬁve groups of sites identiﬁed by the cluster analysis are shown in Table 26.8 with the location of clusters across the gravel bar complex shown in Fig. 26.3. Cluster 1 was found in mid-elevation areas with dry, coarse, poorly sorted sediments and supported sparse pioneer vegetation. Cluster 2 sites were found at the highest elevation and like cluster 1 had coarse, poorly sorted sediments with slightly higher moisture content. The area supported a diverse vegetation community containing a mix of herbs, bryophytes (particularly Racomitrium), and grasses with broom (Sarothamnus scoparius) and gorse (Ulex europaeus) scrub and occasional sycamore saplings (Acer pseudoplatanus). Cluster 3 sites were of mid elevation, very dry and of moderate grain size. They supported a mixed herb-rich mesotrophic grassland. Cluster 4 comprised ﬁner, well-sorted, damp gravels that made up the bar tail areas. They supported a mixed herb/shaded neutral grassland. Cluster 5 contained wet organic low-lying sediments and were found primarily on the lee side of the individual bar tails and where sand splays were developed in the lee of willow bushes. Luxuriant vegetation was prevalent with a composition typical of the margins of palaeochannels in the early stages of abandonment. The overall bar vegetation was dominated by 20 species (e.g., Agrostis canina, Festuca ovina, Viola riviniana, Silene maritimus and Senecio viscos) and accounted for over 70% of the total cover. In total 181 plant species were found of which 87% were noted within the 66 standard sampling units. Plant species richness in samples was relatively high in all clusters with values ranging by almost two fold from 10 in cluster 1–19 in cluster 2. To enable comparison between clusters rarefaction and in one case (cluster 4) extrapolation was therefore used to establish the species pool that would have been associated with each cluster given a constant sampling effort. Table 26.8 reveals that the species pool increases from cluster 1–4 and is lower in 5. Turnover between samples was high in all clusters, although somewhat lower in 2 and 5. The number of unique species was more than twice as high in cluster 2 than the average of the other four clusters. In total 52 species were estimated to be unique to one of the clusters. Overall, the mosaic (with clusters in the ratio observed) supported a species pool of, on average, 1.36 times higher that of an equal sized sample of any one cluster. Although cluster 4 contained almost 92% of the species pool of the whole gravel bar mosaic it had a low uniqueness value compared to cluster 2 with which most of the regionally rare species were associated (Fig. 26.4). This emphasises the importance of sustaining ﬂuvial processes that

D. Gilvear et al.

694

Figure 26.4. Standardised species richness, uniqueness and percentage of the total species pool for the 5 clusters and overall gravel bar mosaic on the River Tummel.

preserve the habitat mosaic if the conservation of similar bar complexes are to be maintained. 3.2.2.

Species– environment relationships

The analysis of species–environment relationships demonstrated that elevation was the most important variable; conﬁrming a number of other studies. Sediment moisture content and surface grain size were also significant in explaining the spatial distribution of the vegetation independent of elevation alone (Fig. 26.5; Table 26.9). Together these represent the most parsimonious set of variables and would be a suitable subset of predictors for subsequent studies at other sites and rivers. Elevation provides a surrogate for one or more deterministic factors (e.g., duration of inundation) and may potentially convey much of the variation in other measured parameters such as soil moisture. However, when the inﬂuence of elevation was tested after ﬁrst ﬁtting all the substrate related parameters it remained highly significant ( p ¼ o0.001).

4. 4.1.

Discussion of case studies Geomorphological controls on plant species recruitment and dynamics

The results of the two studies summarised here, despite being from contrasting European environments, illustrate that elevation and, to a lesser extent, grain size have a significant but differential inﬂuence on the recruitment, survival and development of riparian vegetation. This is supporting the ﬁndings of a number of studies, primarily

Gravel bars

695

Figure 26.5. CCA biplot of species x environmental variables for 66 samples from Ballinluigh Island, R. Tummel. Eigen values axis 1 ¼ 38; axis 2 ¼ 0.20. Species are reduced to initial three letters of genus and species for ease of labeling and some points close to the origin are unlabelled. Species occurring at o5% of sites were excluded from the analysis. Environmental variables are restricted to those that were significant after forward selection followed by Monte Carlo random permutation tests (999 permutations) with Bonferroni correction. The variables shown explained 12% of the variation in species composition.

in the USA and dating back to the early work of Hupp and Osterkamp (1985). Elevation and grain size were major controls on both woody and herbaceous vegetation recruitment and growth, and seral stages of vegetation development. Elevation affects the frequency and duration of inundation and may inﬂuence the likelihood of bed load movement, the latter depending additionally on grain size and shape, as well as patterns of hydrochorous propagule deposition. The significance of hydrochory for seed dispersal has long been recognised (Anderson and Nillson, 1999; Goodson et al., 2002). Plant colonisation of low-lying surfaces is probably facilitated by small ﬂoods spilling across them and conveying vegetative propagules and seeds that are either trapped in the gravels or settle out in their lee. Along the River Tummel, the three bar tails were critical in that they created slack water areas in their lee which allowed the development of vegetation associated with cluster 5. Deposition of ﬁne sediment in these slack water areas was also accentuated by the establishment of dense Salix scrub. Grain size largely controls the amount of available moisture. Although not as significant as elevation, grain size was found to be very important for plant growth of subject species along the Tagliamento. The Tagliamento experiment focused solely on woody species but the Tummel study showed similar ﬁndings in relation to herbaceous vegetation. This ﬁnding when combined with other work in the area suggests that elevation and grain

D. Gilvear et al.

696 Table 26.9.

Species–environmental relationships analysed using CCA.

Variable

Marginal effects TVE

Conditional effects

P

Input order

FVE

F

P

1 5 4 2 3 7 6 ns ns

23.7 8.9 12.6 13.3 13.2 7.2 7.4

3.99 1.60 2.36 2.36 2.28 0.95 1.24

0.001 0.038 0.001 0.001 0.001 0.520 0.140

Elevation Armour sort Armour size % moisture Long axis Sediment size Sediment sort Lateral axis % organic

5.9 4.8 4.6 4.6 4.6 4.4 3.7 2.6 2.2

0.001 0.001 0.001 0.001 0.001 0.001 0.001 0.010 0.038

All variables

21.0

0.001

Note: Marginal effects assess the inﬂuence of each variable, including its covariation with all other variables. The conditional effects consider variables as supplied by a forward selection procedure that ranks variables in accordance with their ability to explain initial and residual variation (see input order). Those variables significant at p ¼ o0.05 based on Monte Carlo random permutation tests (n ¼ 999) and Bonferroni correction are shown in bold. This can be regarded as a minimum adequate model for representing variation in species composition at the site. TVE ¼ % total variance explained; FVE ¼ fraction variance explained by all independent variables.

size may be universally good predictors of vegetation development on gravel bars. However, riparian environments contain a relatively high richness of plant species, which may display notable interspecific variation in response to abiotic conditions (e.g., Francis et al., 2005). Furthermore, the Tagliamento study showed variation in response can be observed based on whether sprouting occurred from seeds or from fragments. This highlights the importance of both interspecific and intraspecific plant variations, which need to be considered alongside the characteristics of the physical environment. On the basis of this study we can propose a conceptual model of the environmental controls on plant diversity on gravel bars. The morphology and the sedimentology of the bar creates spatial heterogeneity and a mosaic of habitats for colonisation and survival. In this context the most important variable is elevation although grain size is also important. The two together control the moisture regime of a patch; sheltered low-lying areas with ﬁne sediments are wetter while coarse sediments at higher elevation have little moisture retaining capacity. Plants once established in low-lying areas are subject to low moisture stress but high disturbance due to frequent inundation, deposition of ﬁnes, and bedload movement; the latter processes opening up areas for pioneer species to establish and improve others. Periodic disturbance of these areas also prevents competitive displacement of pioneer species similar to gap dynamics in forests. Higher up the bar on freely draining gravels, particularly where grain size is large and there is little interstitial sediment only relatively droughttolerant species more typical of coastal shingle or montane scree environments may persist. In such situations, xerophytic species can survive in riverine systems. More

Gravel bars

697

generally, high bar surfaces cause desiccation which may often lead to mortality of young plants in such environments (e.g., Sacchi and Price, 1992; Rood et al., 1998; Johnson, 2000). Consequently, in these areas establishment of deep rooting trees or nitrogen-ﬁxing shrubs provides shade and litter and traps ﬁne sediment thereby aiding moisture retention and soil development that is critical in facilitating establishment of other species. This explains why not only is there a diversity of vegetation composition across the elevational gradient and a high level of species uniqueness at individual levels, but also why at each level turnover between patches is high.

4.2.

Significance for management of riparian corridors

Gravel bars are the main form of sedimentological and elevational variability in gravel-bed rivers and are the key determinant of riparian vegetation development. Plant diversity is supported primarily as a result of high spatial heterogeneity creating a number of distinct ﬂuvially derived habitats for plant colonisation, survival and development. In addition the opposing forces of moisture stress and ﬂuvial disturbance are critical in maintaining species richness and high uniqueness at different elevations by restricting dominance of superior competitors. It is important to consider how human activities can adversely disturb gravel bars and affect their natural functioning and thus their nature conservation value. Change in a river’s ﬂow regime whether natural or anthropogenically driven is likely to significantly affect vegetation via alteration of the nature of geomorphic surfaces, the abundance and composition of dispersing propagules, destruction and survival of seeds and vegetated fragments and bed and vegetation disturbance. Any ﬂow regulation that reduces bedload movement at lower elevations, will for example restrict the regeneration niche of small ruderal species particularly if the regulation of ﬂow also leads to the build up of organic matter which will favour more competitive species. The results of the Tagliamento study suggest that elimination of ﬂoods of a particular size or at a particular time of year by ﬂow regulation would reduce the survival rate of seeds and vegetated fragments by effectively lowering the water table level at bar formations and increasing the negative effects of bar elevation. Importantly some high-level surfaces are only created by rare high-magnitude events (e.g., 1:50 year return period) and if lost due to activities such as channelisation, gravel extraction, or agricultural improvement may prove exceptionally difﬁcult to restore or re-create. Reduction in the size of large ﬂoods will also prevent the creation of high-elevation –coarse-bed material landforms and hence a niche for establishment of more drought-tolerant species. Large, medium and small sized ﬂoods are all therefore critical in maintaining the suite of landforms and processes characteristic of natural gravel bars and islands, and so important to retaining plant diversity. Other studies on the River Tagliamento by Gurnell et al. (2001) and Gurnell and Petts (2002) suggest that the trajectory of biomass production on ﬂuvial surfaces is controlled by the interaction between the chronology of ﬂoods and nature of propagule reproduction. In terms of restoration of river corridors physical placement of exposed riverine sediments and re-sculpturing existing degraded exposures is unlikely to recreate plant diversity without full restoration of the ﬂuvial processes that both

D. Gilvear et al.

698

initiate such features and import potential colonists from the surrounding landscape. This is a challenge in terms of the majority of Alpine and other river catchments draining the mountains of Europe because, natural habitats are fragmented, ﬂows are heavily regulated and adjacent lands often need to be protected from inundation and erosion under existing patterns of land use and to safeguard infrastructure. Overall this study shows that geomorphological processes provide the template for gravel bar vegetation development (Poff and Ward, 1990) and that colonisation and succession of bare gravels, and hence island formation, and their eventual incorporation into the ﬂoodplain is critical in maintaining high ﬂoodplain biodiversity. As such gravel bars and river islands must be regarded as an integral component of functional river corridors and not features that should be removed in the interest of hydraulic efﬁciency. This is the challenge river managers face in the Alpine areas and other rivers draining the mountainous catchments of Europe.

5.

Conclusion

These two case studies demonstrate that the ﬂuvial dynamics of gravel-bed rivers control the nature of plant establishment, survival and development on gravel bars. Morphological diversity is a major factor supporting high species richness, with elevation above normal ﬂow levels being the primary explanatory variable in deﬁning vegetation establishment and development. This must be considered alongside the ecological characteristics and inter- and intraspecific variability of the vegetation. Fortunately there is progress being made in modelling the dynamics of elevation and grain size change in gravel-bed rivers alongside plant response, and thus there is the prospect of coupling biological and geomorphological models to predict the manner in which vegetation may establish and develop on gravel-bed rivers under a range of natural and management scenarios. Overall our ﬁndings are significant in suggesting that gravel bars and river islands are an important component of the river corridor environment and should be left ‘pristine’ where possible. Where they were once present and now absent due to ﬂow regulation or channelisation, consideration should be given to restoring a near natural ﬂow regime and sediment dynamics to allow natural re-establishment of such features in the landscape.

References Abernethy, V.J., Willby, N.J., 1999. Changes along a disturbance gradient in the density and composition of propagule banks in ﬂoodplain aquatic habitats. Plant Ecol. 140, 177–190. Anderson, E., Nillson, C., 1999. Temporal variation in the drift of plant litter and diaspores along a small boreal river. In: Andersson, E. (Ed.), Relationships between Hydrochory and Riparian Flora in Boreal Rivers. Doctoral Dissertation, Umea University, Sweden. Barnes, W.J., 1978. The distribution of ﬂoodplain herbs as inﬂuenced by annual ﬂood elevations. Trans. Wisconsin Acad. Sci. Arts Lett. 66, 254–266.

Gravel bars

699

Bendix, J., Hupp, C.R., 2000. Hydrological and geomorphological impacts on riparian plant communities. Hydrol. Process. 14, 2977–2990. Brice, J.C., 1964. Channel patterns and terraces of the Loup rivers in Nebraska. US Geological Survey Professional paper 422D. Burgess, D., Hendrickson, O.Q., Roy, L., 1990. The importance of initial cutting size for improving the growth performance of Salix alba L. Scand. J. Forest Res. 5, 215–224. Colwell, R.K., 1997. Estimate S: Statistical Estimation of Species Richness and shared species from samples. Version 5.01. University of Connecticut. http://viceroy.eeb.uconn.edu/estimates Dickmann, D., Phipps, H., and Netzer, D., 1980. Cutting diameter inﬂuences early survival and growth of several Populus clones. USDA Forest Service Research Note NC-261. Dykaar, B.B., Wigington, P.J., 2000. Floodplain formation and cottonwood colonisation patterns on the Willamette River, Oregon, USA. Environmental Manage. 25, 87–104. Edwards, P.J., Kollman, J., Gurnell, A.M., et al., 1999. A conceptual model of vegetation dynamics on gravel bars of alarge Alpine river. Wetlands Ecol. Manage. 7, 141–153. Eyre, M.D., Luff, M.L., Phillips, D.A., 2001. The ground beetles (Coleoptera: Caribbidae) of exposed riverine sediments in Scotland and northern England. Biodiversity Conserv. 10, 403–426. Francis, R.A., Gurnell, A.M., Petts, G.E., Edwards, P.J., 2004. Survival and growth response of Populus nigra fragments to differing hydrogeomorphological conditions. In: Webb, B., Acreman, M., Maksimovic, C., Smithers, H., and Kirby, C. (Eds.), Hydrology: Science and Practice for the 21st Century. Volume II. British Hydrological Society, London, UK, pp. 80–89. Francis, R.A., Gurnell, A.M., Petts, G.E., Edwards, P.J., 2005. Survival and growth responses of Populus nigra, Salix elaeagnos and Alnus incana cuttings to varying levels of hydric stress. Forest Ecol. Manage. 210 (1–3), 291–301. Francis, R.A., Gurnell, A.M., Petts, G.E., Edwards, P.J., 2006. Riparian tree establishment on gravel bars: interactions between plant growth strategy and the physical environment. In: Sambrook-Smith, G., Best, J., Bristow, C., and Petts, G.E. (Eds), Braided Rivers: Process, Deposits, Ecology and Management, pp. 361–380. International Association of Sedimentologists Special Publication, 36, 396pp. Gilvear, D.J., Cecil, J., Parsons, H., 2000. Channel change and vegetation diversity on a low-angle alluvial fan: River Feshie, Scotland, Aquatic Conservation. Mar. Freshw. Ecosyst. 10, 53–71. Gilvear, D.J., Willby, N., 2006. Channel dynamics and geomorphic variability as controls on gravel bar vegetation; River Tummel, Scotland. River Res. Appl. 22, 457–474. Girel, J., Manneville, O., 1998. Present plant species richness of plant communities in alpine stream corridors in relation to historical river management. Biol. Conserv. 85, 21–33. Goodson, J.M., Gurnell, A.M., Angold, P.G., Morrissey, I.P., 2002. Riparian seed banks along the River Dove, UK: Their structure and ecological implications. Geomorphology 47, 45–60. Gurnell, A.M., Petts, G.E., 2002. Island dominated landscapes of large ﬂoodplain rivers, a European perspective. Freshw. Biol. 47, 581–600. Gurnell, A.M., Petts, G.E., Hannah, D.M., et al., 2001. Riparian vegetation and island formation along the gravel-bed Fiume Tagliamento, Italy. Earth Surf. Process. Landf. 26, 31–61. Gurnell, A.M., Petts, G.E., Harris, N., et al., 2000. Large wood retention in river channels: The case of the Fiume Tagliamento, Italy. Earth Surf. Process. Landf. 25, 255–275. Hupp, C.R., Osterkamp, W.R., 1985. Bottomland vegetation distribution along Passage Creek, Virginia, in relation to ﬂuvial landforms. Ecology 66, 670–681. Hupp, C.R., Osterkamp, W.R., 1996. Riparian vegetation and ﬂuvial geomorphic processes. Geomorphology 14, 277–295. Jaegii, M.N.R., 1987. Interaction of bed-load transport with bars. In: Thorne, C.R., Bathurst, J.C., and Hey, R.D. (Eds), Sediment Transport in Gravel Bed Rivers. Wiley, Chichester. Johnson, W.C., 2000. Tree recruitment and survival in rivers: Inﬂuence of hydrological processes. Hydrol. Process. 14, 3051–3074. Johnson, W.C., 2002. Riparian vegetation diversity along regulated rivers: Contribution of novel and relic habitats. Freshw. Biol. 47, 749–759. Kalliola, R., Puhakka, M., 1988. River dynamics and vegetation mosaicism: A case study of the River Kamjohka, northernmost Finland. J. Biogeogr. 15, 703–719.

700

D. Gilvear et al.

Karrenberg, S., Kollman, J., Edwards, P.J., et al., 2003. Patterns in woody vegetation along the active zone of a near-natural Alpine river. Basic Appl. Ecol. 4, 157–166. Kollmann, J., Vieli, M., Edwards, P.J., et al., 1999. Interactions between vegetation development and island formation in the Alpine river Tagliamento. Appl. Vegetation Sci. 2, 25–36. Lewin, J., 1996. Floodplain construction and erosion. In: Petts, G.E. and Calow, P. (Eds), River Flows and Channel Forms. Blackwell Science Ltd., Oxford, pp. 203–220. Marston, R.A., Girel, J., Patou, G., et al., 1995. Channel metamorphosis, ﬂoodplain disturbance and vegetation development, Ain River, France. Geomorphology 13, 121–132. Moreton, D.J., Ashworth, P.J., Best, J.L., 2002. The physical scale modelling of braided alluvial architecture and estimation of subsurface permeability. Basin Res. 14, 265–285. Nanson, G.C., Croke, J.C., 1992. A genetic classiﬁcation of ﬂoodplains. Geomorphology 4, 459–486. Parsons, H., Gilvear, D.J., 2002. Valley ﬂoor landscape change following almost 100 years of ﬂood embankment abandonment on a wandering gravel-bed river. River Res. Appl. 18, 461–479. Petts, G.E., Gurnell, A.M., Gerrard, A.J., et al., 2000. Longitudinal variations in exposed riverine sediments: A context for the ecology of the Fiume Tagliamento, Italy. Aquat. Conserv. Mar. Freshw. Ecosyst. 10, 249–266. Poff, N.L., Ward, J.V., 1990. Physical habitat template of lotic systems: Recovery in the context of historical pattern of spatiotemporal heterogeneity. Environ. Manage. 14, 629–645. Prach, K., 1994. Vegetation succession on river gravel bars across the northwestern Himalayas, India. Arctic Alpine Res. 26, 117–125. Richards, K., Brassington, J., Hughes, F., 2003. Geomorphic dynamics of ﬂoodplains: Ecological implications and a potential modelling strategy. Freshw. Biol. 47, 559–579. Robertson, K.M., Augspurger, C.K., 1999. Geomorphic processes and spatial patterns of primary forest succession on the Bogue Chitto River, U.S.A. J. Ecol. 87, 1052–1063. Rood, S.B., Kalischuck, A.R., Mahoney, J.M., 1998. Initial cottonwood seedling recriutment following the ﬂood of the century of the Oldman River, Alberta, Canada. Wetlands 18, 557–570. Sacchi, C.F., Price, P.W., 1992. The relative roles of abiotic and biotic factors in seedling demongraphy of arroyo willow (Salix lasiolepis). Am. J. Botany 79, 395–405. Sadlar, J.P., Bell, D., Fowles, A., 2004. The hydroecological controls and conservation value of beetles on exposed riverine sediments in England and Wales. Biol. Conserv. 118, 41–65. Scott, M.L., Shaforth, P.B., Auble, G.T., 1999. Responses of riparian cottonwoods to alluvial water table declines. Environment. Manage. 23, 347–358. Shin, N., Nakamura, F., 2005. Effects of ﬂuvial geomorphology on riparian tree species in Rekifune River, northern Japan. Plant Ecol. 178, 15–28. Simons, D.B., Simons, R.K., 1987. Differences between gravel and sand bed rivers. In: Thorne, C.R., Bathurst, J.C., and Hey, R.D. (Eds), Sediment Transport in Gravel Bed Rivers. Wiley, Chichester. Steiger, J., Gurnell, A.M., 2003. Spatial hydro-geomorphological inﬂuences on riparian zone sedimentation: Observations from the Garonne River, France. Geomorphology 49, 1–23. Tabbacchi, E., Planty-Tabbacchi, A.M., Salinas, M.J., et al., 1996. Landscape structure and diversity in riparian plant communities: A longitudinal comparitive study. Regulated Rivers Res. Manage. 12, 367–390. Tockner, K., Ward, J.V., Arscott, D.B., et al., 2003. The Tagliamento River: A model ecosystem of European importance. Aquat. Sci. 65 (3), 239–253. Van Coller, A.L., Rogers, K.H., Heritage, G.L., 2000. Riparian vegetation–environment relationships: Complimentarity of gradients versus patch hierarchy approaches. J. Vegetat. Sci. 11 (3), 337–350. Ward, J.V., Tockner, K., Ueginlinger, U., et al., 2001. Understainding natural patterns and processes in river corridors as the basis for effective restoration 2001. Biodiversity: Towards a unifying theme for river ecology. Regulated Rivers Res. Manage. 17, 311–323.

Gravel-Bed Rivers VI: From Process Understanding to River Restoration H. Habersack, H. Pie´gay, M. Rinaldi, Editors r 2008 Elsevier B.V. All rights reserved.

703

27 River restoration in the Alps and their surroundings: past experience and future challenges Helmut Habersack and Herve´ Pie´gay

Abstract Alpine rivers have undergone signiﬁcant changes over the two last centuries. Human activities have modiﬁed their geometry through engineering measures to gain land for agricultural purposes and settlements, as well as through active mining to exploit gravel resources. Their sediment and water transfers have also been altered by hydropower-plant construction, control works on high-gradient streams, and catchment land-use changes. The resulting river morphological changes have led to abiotic (e.g., river-bed degradation and narrowing) and biotic (e.g., longitudinal and lateral disconnection) disruption. The current critical management situation (channel instability problems, ﬂood effects, biodiversity decrease) has made river restoration a major issue in the Alps and their surroundings. Such an approach is reinforced by the European Water Framework Directive, which aims to ensure that rivers attain a good ecological status by 2015. In the Alps, space is not always easily available and boundary conditions have changed over the long term. A major challenge in river restoration in the Alpine environment is therefore to identify the processes and key parameters for improving both geomorphological and ecological conditions under often-restricted boundary conditions. Early attempts at river restoration mainly focused on small-scale measures. Today, successful restoration projects in high-energy and bedload-transportdominated conditions must include the full spectrum of scales, striving to initiate self-forming morphodynamics. In this context, we appraise restoration experiences from the Alps, focusing on channel widening and dike enlargement, former channel reconstruction and reconnection, promotion of bedload supply input from ﬂoodplains, tributaries, and hillslopes, as well as on bank erosion measures and restoration activities. We discuss the basic arguments behind such actions, their limitations, and research challenges.

E-mail address: [email protected] (H. Habersack), [email protected] (H. Pie´gay) ISSN: 0928-2025

DOI: 10.1016/S0928-2025(07)11161-5

H. Habersack, H. Pie´gay

704 1.

Introduction

In Europe, all large rivers have been signiﬁcantly modiﬁed by human activities for centuries (Petts, 1989). A study of the International Commission for the Protection of the Alps (CIPRA) showed that only around 10% of the most important rivers of the entire Alpine region are still ‘‘pristine’’ or in a ‘‘near-natural’’ condition (Martinet and Dubost, 1992). The earlier activities concentrated on local works for ﬂood and bank erosion protection but also for navigation purposes. The basic river morphological features underwent serious disruptions in the 19th century, when systematic hydraulic engineering measures were conducted along reaches of 1 to more than 100 km length. These major and widespread regulation efforts totally modiﬁed the morphodynamics and sediment transport behaviour, initiating channel changes such as river-bed degradation (Habersack and Nachtnebel, 1995; Bravard et al., 1997). These morphological changes led to abiotic and biotic consequences, and are associated with ecological and economic impacts (Roux et al., 1989; Bravard et al., 1999a; Jungwirth et al., 2003). Today’s critical situation in terms of management (channel instability problems, limitations of ﬂood regulation works, biodiversity decrease) has made river restoration a major issue in the Alps and their surrounding areas (Bravard et al., 1999a; Gilvear, 1999). These efforts are reinforced by the European Water Framework Directive (WFD), which aims to ensure that rivers will reach good ecological status by 2015 (European Parliament and Council of Europe, 2000). The ﬁrst step in addressing these concerns is to describe the existing situation. This incorporates a deﬁcit analysis, which highlights the risk that a certain water body will not reach a good ecological status by 2015. This is followed by a planning phase. In the ﬁnal phase, measures are implemented to reach the goal. To date, no formal guidelines for restoration are available to achieve these goals. This calls for a summary of current measures in order to evaluate the most appropriate applications. In the Alpine environment, high gradients generate signiﬁcant energy and channel features are sensitive to changes in control parameters (peak ﬂow regime and bedload input). A key challenge in river restoration is to identify the processes and primary parameters with which to improve both geomorphological and ecological conditions, under often-restricted boundary conditions (Ja¨ggi, 1989; Bravard et al., 1999b; Pie´gay and Schumm, 2003). Early attempts at river restoration mainly applied small-scale measures. However, successful restoration projects in high-energy and bedload-transport-dominated conditions have to emphasise the full spectrum of scales, striving to initiate self-forming morphodynamics (Amoros et al., 1982; Frissell et al., 1986; Amoros and Petts, 1993; Habersack and Nachtnebel, 1995; Habersack et al., 2000). In this context, we present the historical evolution of rivers in the Alpine area and their surroundings, and discuss how this evolution has led to the necessity of river restoration. A summary of restoration experiences is then drawn. Feedback elements are provided to explain such actions and their limitations, highlighting research challenges.

River restoration in the Alps and their surroundings 2.

705

Problem statement: the human impact in an Alpine context

2.1.

Short history of river regulation: from natural to cultural landscapes

Within the Modern period the ﬁrst signiﬁcant measures were taken on the large rivers to prevent ﬂoods (with dikes) and to improve navigation (Table 27.1). Industrial development in the 19th century, along with increased technical capabilities and public investment, brought about major channelisation works on the Rhine, the Rhoˆne, and the Danube. The goal was to develop navigation and protect major cities from ﬂoods (Vischer, 1986, 2003). Moreover, the main tributaries of these rivers were also embanked in order to promote agriculture in the Alpine valleys (Bravard and Peiry, 1993; Girel et al., 1997). Most of the streams were also controlled by farmers, who used riparian wood, ploughed the lands, and protected the banks with traditional techniques (mainly bioengineering methods, Kondolf et al., 2007). The channelisation work, which initially began on navigable rivers, was then extensively applied on their tributaries (Fig. 27.1). Moreover, agricultural development and expanding industrial areas exerted great pressure on ﬂoodplain areas. This was accompanied by bank protection measures and forest clearance. Beyond this, a high demand for gravel for infrastructure construction and buildings led to major gravel mining in rivers between the 1960s and 1980s (Habersack et al., 2000). In the late 19th century and then after the Second World War, major impacts resulted also from damming for hydroelectricity production and improved ﬂood protection (the Rhine in the 1960s, the Rhoˆne up until the 1980s, the Danube up until today). The large rivers were particularly important because their discharge enabled continuous electricity production. At the same time, their catchments, with high slopes and important water resources availability, were also intensively regulated to provide electricity during peak periods. This led to important water storage efforts and inter-basin transfers (Edouard and Vivian, 1984; Peiry and Marnezy, 2000). Table 27.1.

Overview of historical river engineering measures at large European rivers.

River

First engineering measures

Systematic hydraulic engineering measures at long river reaches

Po

Embankments 13th–14th centuries

Rhoˆne

18th––early 19th

French Alpine large rivers Rhine

16th–17th (embankments for protecting the towns) Channelisation, bank protection 18th century Flood retention 14th–16th century Improvement for shipway 15th century

Mainly 18th century downstream of Piacenza 1876–1884 (Ing. Jacquet) 1884–1920 (Ing. Girardon) Ise`re (1829–1845); Arve (1820–1838); Var (1844–1869) Beginning 1804 (Tulla)

Elbe Danube

1821–1905 1830–1890

Source: Garbrecht (1985); Vischer (1986); Braga and Gervasoni (1989); Bravard and Peiry (1993); Tricart and Bravard (1991); Poinsart and Salvador (1993).

H. Habersack, H. Pie´gay

706

(a)

Existing situation

1817 8 17 11876 1 8 7 6

Flow direction

1 km

6000

slope 1876 - today

radius 1817-1876-today 1817

today

3

4000

2000 1400

135

125

1

115

105

1

95

2

kilometers

1000

width [m]

2

0

12000

800

today

1876

1876

slope [‰]

radius [m]

(b)

3

reach

4

5

width 1817 - 1876 - today max. 1817 mean 1817 max.1876 mean 1876 max.today mean today

600 400 200 0 1

2

3 section

4

5

Figure 27.1. (a) Channelisation of the Mur River in Styria/Austria (reach 3 on the map). Sections 1–5 according to geomorphic characteristics 1: straight, 2: gorges, 3: historically braided, 4: constrained in urban areas, 5: former braided-meandering; (b) changes of morphological parameters over time (width, slope, radius) at the Mur (after Habersack and Schneider, 2000, permission obtained from Wasser & Boden).

River restoration in the Alps and their surroundings

707

This history of regulation was reinforced by human activity within the catchments. At the end of the 19th century, foresters restored the hillslopes that had been destabilised due to overgrazing and deforestation throughout the 18th and 19th centuries as a result of demographic pressure. This approach was taken in France with the RTM services (Restoration of Mountain Terrains), but also in Bavaria, Austria, and Italy. Different measures were implemented, such as the afforestation of slopes along with the construction of check-dams to break the slope of torrents and stabilise the valley sides. These efforts helped to store the bedload sediment, enabling it to be easily removed. All this work also contributed to modifying the sediment input in the channel network, sometimes initiating channel metamorphosis (Pie´gay and Salvador, 1997). In downstream areas this planned afforestation work was strongly reinforced by major land-use changes in the mountain area. Thousands of hectares of ploughed and grazed land were abandoned by farmers migrating to the cities, leading to major spontaneous afforestation between 1930s and 1970s in the French Pre-Alps. This, in turn, increased the sediment trapping capacity of the catchments and provided a longterm sediment deﬁcit downstream, as shown on the Droˆme (Lie´bault and Pie´gay, 2002). In Austria, afforestation also played an important role, signiﬁcantly inﬂuencing the sediment regime and related river morphology (Habersack et al., 2003). As a consequence, by the late 20th century the channel network across the Alps region was strongly regulated (Table 27.2), and ﬂow and sediment regimes were strongly modiﬁed. The sediment transfer has been reduced by disconnecting the sources from the channel network but also by sediment removal and intermediate storage. The lack of bedload, together with other hydraulic modiﬁcations, lead to river-bed degradation and, ultimately, a larger width-to-depth ratio. This is the dominant situation of Alpine rivers, which are generally supply-limited systems. The cumulated length of braided reaches decreased by 70% in France (Fig. 27.2, Bourdin, 2004) and 95% in Austria (Muhar et al., this volume) during the 20th century. In rare cases, bedload surpluses exist (reservoir siltation, but also upper reaches still directly connected to active, bedload sources). This leads to enhanced morphodynamics and eventually to a reduced width-to-depth ratio, but also channel aggradation and associated ﬂooding risks in highly humanised valleys.

2.2.

River responses to regulation and coping with new demands and needs

Before the 19th century, local engineering measures did not totally change river morphology; sediment input, transport, and output were still the same and, in general, characteristic channel features were preserved. Over the short term, the main goals of systematic river engineering measures in the 19th and 20th century were achieved. This included improved shipways, ﬂood protection and transport capacity, reduced channel aggradation, and increased availability of useable agricultural land. Toward the late 1970s, however, ecological deﬁcits showed the negative effects of the anthropogenic measures, prompting a number of small-scale river restoration measures (Jungwirth et al., 1993). Then, over the last 10–20 years, these ecological deﬁcits were supplemented by crucial physical problems, especially with respect to river morphology (long-term development, Surian and Rinaldi, 2003), water resources,

Countries Peak ﬂow lowering by upstream dam

X

X

Bank protections Embankment/ Discharge Riparian to prevent impact on lowering in a by- encroachment in channel ﬂooding passed channel the active bars movement

Groundwater lowering and riparian dewatering

X X

X X

X X X

X X X

X X X X X

X X X

X X

X

X

X X

X

X X

X X X X X X X X X X

X X

X X X X

X X

X X X X X X X X X X X

X X X X X X X X X X X X X X X X

X X X X X X X X X X X X X X X

X X

X

X X X X X

Note: G, Germany; A, Austria; F, France; CH, Switzerland; I, Italy; rivers selected based on existing restoration projects. Acknowledgement: M. Rinaldi, D. Sogni, N. Surian (Italy); M. Jaeggi, B. Zarn (Switzerland); Schaipp (Bavaria); J.N. Gautier (France).

X X X X X X X X X X

H. Habersack, H. Pie´gay

Rhine River G–F Upper Rhine River A–CH Rhoˆne River F Arve River F Droˆme River F Durance River F Ise`re River F Doubs River F Ain River F Danube River G–A Drau River A Sulm River A Lech River G–A GroXache River A Gail River A Salzach River G–A Isar River G Thur River CH Emme River CH Tagliamento River I Piave River I Brenta River I Gesso River I Adda River I Dora Baltea River I

Upstream In-channel sediment works trapping provoking degradation (mining, groins for navigation)

708

Table 27.2. Main impacts on processes and channel structure at selected rivers located in the European Alps and their surrounding areas during the last two centuries.

River restoration in the Alps and their surroundings End of 20th century

mid 19th century

Braided reaches

709

FRANCE

20 km

Figure 27.2. Length of braided rivers in the French Alps in the mid-19th century before the major regulation works (estimated to be 1100 km) and in the late 20th century (estimated to be 650 km) (after Bourdin, 2004).

and safety conditions. Various negative consequences of river and basin regulation policy were progressively recognised in the Alps and the surrounding areas:

Regressive and progressive erosion downstream of torrent control structures, reservoirs, and gravel mining sites; progressive erosion downstream of dams, with groundwater lowering, undermining, and failures of dikes and bridges; bed degradation following slope increases (e.g., meander cut-offs), decrease of river bed width (increasing shear stress), blockage of side erosion by bank protection measures, with consequences for human safety and water resources preservation (Ja¨ggi and Zarn, 1999). Channel narrowing due to peak ﬂow reductions downstream of certain dams, with associated ﬂood damage consequences. Some reservoirs only modify current ﬂood peaks. Human settlements may suffer ﬂooding problems during large ﬂoods, especially in downstream reaches. This reﬂects narrower and rougher channels downstream as well as greater water depth (compared to pre-construction conditions) in response to vegetation encroachment. Ecological consequences of river engineering measures: reduced biodiversity of ﬂoodplain and aquatic ecosystems (impact on ﬁsh migration, habitat diversity decrease, etc., see Bravard et al., 1986; Wasson et al., 1998; Jungwirth et al., 2003).

The most striking effects were the severe loss of natural habitats. This led to a massive decline in plant and animal diversity, and even to the extinction of many

H. Habersack, H. Pie´gay

710

species (Rohde, 2004). In Switzerland, e.g., the ﬁsh species salmon (Salmo salar), sturgeon (Acipenser sturio), and lamprey (Lampetra fluvialis) were lost (BUWAL, 2002). In the 1980s, scientists ﬁrst began to understand the long-term consequences of such modiﬁcations (Jungwirth, 1986; Ja¨ggi, 1989, Table 27.2). It became clear that dams, embankments, and mining activities in rivers had additional long-term costs, and that the efﬁciency of engineering solutions was limited in certain cost-beneﬁt contexts (Pie´gay et al., 1997). The 1990s marked the emergence of the sustainable development perspective and the notion of ‘living with nature’. This led to new concepts based on the idea that preserving nature was a good strategy because natural functions provided human beneﬁts. The concept of buffer strips to prevent water quality degradation in agricultural basins is one of the key arguments for renaturating river margins (Pinay et al., 1994; Qiu and Prato, 1998). New strategies to manage ﬂoods and bank erosion, i.e., giving rivers more space, are also widely accepted amongst managers (Dister, 1992; Pie´gay et al., 2005). River restoration is called upon to integrate the whole catchment into a sustainable entity (Downs and Gregory, 2004).

2.3. 2.3.1.

Some examples of human impact Hydropower plants and engineering measures in Austria

In Austria, the hydrological impacts of hydropower plants signiﬁcantly altered rivers (Muhar et al., 1998; Jungwirth et al., 2003). Hydropower plants interrupted the longitudinal river ﬂux, both biologically and concerning the sediment continuum, causing a deﬁcit downstream. Sixty-percent of the Austrian rivers are impacted by water abstraction and subsequent residual ﬂow, surge effects, impoundments, discontinuum due to hydraulic structures, and deﬁcits in river morphology (Lebensministerium, 2005). The river morphological deﬁcits are related to sediment transport modiﬁcations following torrent control, river engineering measures, and impoundments. Sediment deﬁcits along with reduced river-bed width, increased slope, and blocked side erosion triggered a demand for river engineering and restoration measures. Almost the entire length of the large rivers in Austria shows either a sediment deﬁcit or surplus (Table 27.3). Sediment deﬁcits and increased transport capacity due to channelisation promote bed degradation. The degradation of some Austrian rivers reaches 1–6 cm per year; the Salzach River (border section to Germany) has already experienced a degradation of over 7–8 m (Stephan et al., 2003). On the other hand, the gravel layer (Quaternary deposits) is relatively thin, and the river bed will at some point in the near future reach the underlying ﬁne-grained marine (Tertiary) deposits. At the Salzach River, this already occurred once in the 1960s in the form of a riverbed ‘‘breakthrough’’ (German section, Fig. 27.3). In such cases, all the water at low stage ﬂows in a narrow, canyon-type channel. This leads to technical and ecological problems. In the upstream border section between Austria and Germany, the 100-years ﬂood of 2002 caused a new river breakthrough (Fig. 27.3). Numerical simulations had predicted before the ﬂood that, assuming regular hydrological

River restoration in the Alps and their surroundings Table 27.3.

711

Deficit and excess of sediments for large Austrian rivers.

River

Gauging station

Mean discharge (m3s1)

Danube Drau Mur Enns Inn Salzach

Wien Reichsbr. 1931 Drauhofen 112 Bruck/Mur 105 Steyr 202 Kirchbichl 289 Golling 142

Analysed length (km)

Deficit (% length)

Excess (% length)

350 214 280 186 258 182

30 27 66 33 44 49

70 59 24 47 29 13

conditions, the gravel layer should have been sufﬁcient for 10 years. The single ﬂood in 2002, however, caused the breakthrough. As a consequence, river-bed widening – the originally suggested restoration measure – is hardly suitable anymore (due to lack of gravel) and other options (e.g., ramps, another hydropower plant) are being discussed. The danger of bed breakthrough is also evident for other Austrian rivers such as the Danube downstream of Vienna and the Mur River at the border with Slovenia. Note that monitoring results based on the EU Water Framework Directive might have still shown the Salzach river reach to have a good ecological status with respect to ﬁsh biology before the 2002 ﬂood. Immediately after the 2002 ﬂood-related bed breakthrough, however, the good ecological status was lost along these strongly affected reaches. This demonstrates that the morphodynamic changes of rivers must also be incorporated in future monitoring concepts that have ecological aims. Pure biological parameters can only reﬂect the momentary condition. 2.3.2.

In-channel mining in France

Mining activities in river channels began along most of the main branches in the 1960s, but activity peaked in the 1970s to early 1980s. Such mining was ﬁnally forbidden in 1994. This activity was originally promoted because of the high quality and high resource availability (large active channel). The low cost of extraction near urban areas and chief transportation routes along main valleys, mainly for the construction of the highway network, was also a plus. Moreover, the activities were thought to contribute to ﬂood management protection by lowering the bed and increasing ﬂow capacity. State agencies usually authorised the extraction of a volume corresponding to the ‘‘annual bedload transport’’. Unfortunately, the calculation was based on hydraulic formulae that often over-estimated the bedload transport rate. Finally, the mining companies tended to surpass the authorised volume because the impacts of such activities were poorly understood and the public thought it was a positive approach to prevent ﬂooding. Such mining activities created unexpected problems. (i) In the inner Alps, the gravel layer was thin, and regressive and progressive erosion rapidly excavated through to lacustrine deposits (mainly sand, silt, and peat). This induced very rapid and widespread degradation by 12–14 m in the northern Alps (Fier and Arve Rivers) (see Peiry, 1987; Peiry et al., 1994). (ii) Degradation was more severe than expected along most of

H. Habersack, H. Pie´gay

712 (b)

(a)

407 406 (c) 405

Degradation 1953 to 2001: ca. 3 m

404

Degradation flood August 2002: 3 - 4 m, 2 pools

403 m a. sl.

402 7.10.53

401 400

18.03.75 ca. 3m

399

10.10.95

398

gravel

397 15.10.01

396

ca. 3m

fine, marine ca. 4m sediments

395 26.11.02

394 393 0

10

20

30 40

50

60 70 80 distance [m]

90 100 110 120 130 140

Figure 27.3. Bed ‘‘breakthrough’’ at the Salzach River (a) in the 1960s (Hengl, 2004); (b) Schematic drawing of top gravel layer, forming the gravel bed, and lower Tertiary, ﬁne material layer (Hengl, 2004); (c) during the 100-year ﬂood in 2002 (Modiﬁed after WWA Traunstein, 2004).

the branches, undermining local infrastructures. Numerous bridges collapsed and kilometres of dikes built during the 19th century were destabilised. In the Ubaye River, where extraction took place at the conﬂuence with the river and one of its main tributaries, 5 m of degradation at the mining site was observed in 1986 and regressive erosion upstream destabilised the embanked reach. (iii) Water resource availability was disrupted along some rivers because the demand for water pumping increased after wells dried due to lowered water tables. (iv) Within the aquatic part of the channel, simpliﬁcation of the habitat conditions (rock outcropping, temperature

Surface area of the aquatic zones of the former channels (centred and normed values)

River restoration in the Alps and their surroundings

713

1.5 1 1953

1969

1976

Studied aquatic zones

0.5

ingm mera mmer gri tra cru layn lays var ingt lon char vpl

1978

0 -0.5 1991

-1

1996

1989

-1.5 -2 0

50

100

150

200

250

300

Daily discharge of the Doubs (m3/ s) Figure 27.4. Evolution of the former channel lakes along the Doubs River between 1953 and 1996 in relation to daily discharge. Whereas the 1953 and 1996 observations correspond to a similar discharge, the aquatic areas of the former channel have signiﬁcantly decreased, mainly in the 1970s in relation with inchannel gravel mining activity (after Citterio et al., 2002). The discharge of 1953 is not known but there is a 95% probability that it is in the range indicated by the horizontal arrows according to the known discharges for this period of the year.

increase in the water column and lower gradient between control gradient weirs, homogenised water depth and velocity) had ecological consequences. Communities were also modiﬁed in the riparian areas, where the groundwater table declined and in some cases created drier conditions with intense riparian tree mortality (Bornette and Heiler, 1994). Along the Doubs River, the history of mining activity was reconstructed and the evolution of the former channel areas from 1953 to 1996 surveyed. These former channels are interesting ecosystems worthy of preservation. A clear decrease in aquatic area of the former channels in the 1970s was observed, mainly in the reaches where extraction occurred (Fig. 27.4). The problems associated with mining remain unresolved. In the Droˆme River, 10 years after mining activity stopped, a bridge collapsed in December 2003 due to undermining. The question of dike restoration is now being posed to local elected ofﬁcials in order to determine future ﬂood control strategies in urban areas. On the Doubs River, the pits are still active and continue to trap sediments and induce progressive erosion downstream.

2.4.

Conclusions

Almost all Alpine rivers have been anthropogenically inﬂuenced, with consequences reaching from short-term hydrological to long-term sediment- and morphology-related

H. Habersack, H. Pie´gay

714

impacts. The resulting ecological, technical, and economic problems require rethinking past and current river management policies. Successful river restoration integrates ecological needs, morphological goals, ﬂood protection, water resources, recreation, and landscape aspects. The goal is to live and work with nature rather than to ﬁght it. This requires a complete behaviour change in earth resource management and management programmes, embracing the new demands that emerged in the late 20th century. The rivers and their immediate surroundings are attractive sites for leisure activities, and the public now demands preserving and improving the living environment and landscapes. This historical evolution opens new avenues for restoration concepts, mitigation, and rehabilitation measures.

3. 3.1.

Currently applied restoration measures Short historical review

The ﬁrst restoration measures aimed mainly at restructuring short river reaches, especially by increasing variability in channel width and depth or designing certain habitats such as spawning places and gravel bars. Examples of initial attempts to restore rivers include reintroduction of gravel along Danish streams (Brookes, 1990), remeandering channels – but with stabilised curves – in different areas of Europe (Binder et al., 1983; Brookes, 1987), and reconnecting former channels to the main channel (Glitz, 1983; Mo¨ller and Wefers, 1983). In Austria in the 1980s, a variety of small-scale restoration projects were realised (Jungwirth et al., 1993), focusing on restructuring the river channel by means of groynes, bioengineering measures, and designing the low water course and river banks. These measures were important as pilot projects for river restoration but are not always sustainable, e.g., a single larger ﬂood could destroy the original restoration measures by channel aggradation, degradation, or erosion. The early 1990s marked the ﬁrst scientiﬁc debates on restoration and self-restoration processes (Jungwirth et al., 1993; Sear, 1994; Henry and Amoros, 1995a; Kondolf and Downs, 1996; Rutherfurd et al., 1998). This was followed by debate on the effectiveness of local projects versus slight but potentially more efﬁcient modiﬁcations in management policy (Iversen et al., 1993). Some scientists and managers argued that while local projects are hotspots in which one can communicate and demonstrate the efﬁciency of restoration propositions, they are very expensive and their cost-effectiveness has to be analysed. Reconsidering the channel processes in restoration attempts also proved to be crucial. Bank erosion is a case in point: this process is often considered negatively because it is a natural hazard, whereas it is recognised as important in preserving ecosystem diversity. The same evolution can be observed with regard to ﬂooding: new concepts are emerging to manage ﬂoodwater at the basin scale by creating or recreating ﬂoodplains to prevent major ﬂooding downstream (ﬂoods in Central Europe in 2002, ﬂood within the Rhoˆne delta in December 2003). The idea that gravel and bedload transport is equally important for river ecosystem

River restoration in the Alps and their surroundings

715

functionality and technical demands (e.g., bank stability) is gaining widespread acceptance. With the Water Framework Directive, river managers who previously considered river alteration mainly in terms of water chemistry are now called upon to assess aquatic ecosystem quality and appreciate both biological and hydromorphological quality. This new perspective underlines a new way of thinking and the opportunity to consider physical restoration in environmental planning and management strategies. It also opens exciting possibilities to share national experiences and develop network expertise.

3.2. 3.2.1.

Restoration projects around the Alps General overview

Most restoration projects in the Alps and their surroundings consider the related natural processes, but in various detail (Table 27.4). Measures depend primarily on the human impacts along a given reach. In principal, the restoration measures act on ﬂuxes (mainly the ﬂow regime) or on the channel structure to improve the interactions between ﬂoodplain and channel or the aquatic habitat conditions. Actions aimed at sediment ﬂux are still rare because they are less manageable than ﬂows, which are modiﬁed by dams and which can still be improved by dam management. Sediment ﬂux is a more complicated issue because sediments are trapped further upstream and timeframes of response are considerably longer. To date, only few pioneering restoration projects deal with sediment transport and related river morphology in the Alps (Table 27.4). Concerning ﬂow, increasing the minimum ﬂow downstream of dams is already a common practice in Austria, Bavaria, France, and Italy. On the other hand, reducing surge effects caused by hydropower plants remains a major challenge for the coming years, especially within the Water Framework Directive. Morphologically optimised restoration projects may even fail to yield a good ecological status due to surge effects (Unfer et al., 2004; Muhar et al., this volume). Recent strategies attempt to improve both ﬂood protection and ecosystems. In this context, channel widening is a common approach in Switzerland and Austria (Ja¨ggi and Zarn, 1999; Habersack et al., 2000). Opening dikes and re-ﬂooding certain sections of ﬂoodplains is also gaining importance (e.g., Rhine River). Actions on former channels are a key issue in structural restoration because this approach improves the quality of riverine ecosystems (Klein et al., 1994; Henry and Amoros, 1995a,b; Schiemer et al., 1999; Amoros, 2001; Amoros et al., 2005). Old channels were typical features along the peri-alpine rivers prior to regulation, and their restoration is a well-accepted strategy by local authorities. Until now, alluvial forests have been somewhat neglected in restoration efforts, partly because the social constraints are higher. This reﬂects issues such as the need to convert agricultural areas. In addition the public image of forests in terms of natural heritage is weaker than that of water bodies, especially when hillslope forests are already widespread and well preserved.

716

Table 27.4.

Restoration options at selected rivers located in the European Alps and their surrounding areas. Open or remove dikes for ﬂooding

Create an River bed In-stream bed Increase erodible widening structure minimum corridor for improvement ﬂow preserving bank erosion X

Restoration Artiﬁcial Former of river gravel channel continuum resupplying restoration biological (B) sediment transport (S) B B S

X X

Raise the groundwater level (by ramp or other options)

Sustainable removal of vegetation established in the active channel

X X

X X X

X

X X

X X

X X X X X X X X X X

X X X X X X

S B

X X

X X X

S

X

X

B, S B, S

X

X

X

X

X

X

X X

X

Acknowledgement: M. Rinaldi, D. Sogni, N. Surian (Italy); M. Jaeggi, B. Zarn (Switzerland); Schaipp (Bavaria); J.N. Gautier (France).

H. Habersack, H. Pie´gay

Rhine River X Rhoˆne River Arve River Droˆme River Durance River Ise`re River Doubs River Ain River Danube River Drau River Sulm River Lech River GroXache Gail River Salzach River Isar River Thur River Emme River Tagliamento River Piave River Gesso River

Remove dike or bank protection to reactivate erosion

River restoration in the Alps and their surroundings

717

Restoration performed primarily for ecological purposes is atypical. Measures are generally integrated in an overall strategy of basin or reach management that applies the concept of sustainable development. On large rivers, complex projects usually combine ‘‘renaturation’’ and management improvement to satisfy immediate needs. Multi-functional approaches such as along the Austrian Danube demonstrate useful strategies in the context of future restoration of large rivers (Tockner et al., 1998). The Danube downstream of Vienna was once a large partially braided system. Today, it is regulated and inﬂuenced by an upstream hydropower chain, causing bedload deﬁcit. It also now forms the core of a National Park, which must reconcile many issues such as a signiﬁcant river-bed degradation tendency (3–4 cm per year), an international shipping route, and maintaining ﬂood protection. Future restoration measures are being discussed in an integrated project designed to simultaneously improve the morphological situation (stopping river bed degradation), maintain the shipway (minimum water depths during low discharges), as well as achieve ecological functionality (National Park, side arm reconnection, side erosion; Reckendorfer et al., 2005, Habersack et al., 2007). 3.2.2.

Act on channel geometry: the example of channel widening

Channel widening has become a very common restoration measure in Austria and Switzerland (Ja¨ggi and Zarn, 1999; Rohde, 2004). According to Hunzinger (1998) the morphological changes along the longitudinal proﬁle of a river widening are:

The mean bed level in the widening is stepped vertically relative to the bed level in the upstream and downstream channel to ensure continuity and energy conservation. A new equilibrium slope becomes established. This is steeper than the slope of the original narrow streamway. In the case of long river widenings, this effect increases the upstream bed level. Bars are formed, creating a more diverse ﬂow pattern. At the same time, cross ﬂows and scouring lead to an increased hydraulic load on the river banks. The ﬂow is concentrated in the downstream part, causing intense scouring at the constriction. Sediment is retained within the widening, causing temporary downstream erosion.

River-bed widening is one of the key issues in restoring river morphology, as major changes during river channelisation brought about a decrease in river-bed width. At the Drau River, e.g., problems with ﬂood protection, river bed degradation (Habersack and Nachtnebel, 1998), and ecology have been appraised through an interdisciplinary project. Among the considered alternatives, those reducing transport capacity or shear stresses by widening the river channel have yielded positive results. Further measures of interest are those that increase bedload input either from tributaries or through side (lateral) erosion. Restoration measures (mainly river-bed widenings) have been realised during and after the project ‘‘stream care scheme Upper Drau’’ since 1991 and especially since 2001 within the EU LIFE project ‘‘Restoration of the wetland and riparian area at the Upper Drau River, 1999–2003’’ (currently about 10 of ca. 70 km are restored); the latter has now been extended in a

718

H. Habersack, H. Pie´gay

Life II project for another 3 years. One of the most recent and dynamic examples is Kleblach/Lind (2 km long, channel width 40–50 m, bed slope about 0.0015 mm1, 100-years ﬂood about 1000 m3s1, mean grain size around 13 mm; Habersack and Laronne, 2001). At Kleblach/Lind a side channel was initiated along the left bank of the Drau River (Fig. 27.5). In this project, river-bed width was doubled in the lower and upper part of the section (Habersack et al., 2003). The main goals were to: (i) stabilise the river bed by increasing the bed width, (ii) initiate natural morphological developments, (iii) initiate improved habitats for plants and animals, and (iv) improve ﬂood protection. An intensive monitoring program was developed to check the achievement of deﬁned goals. Comparisons of the river geometry showed that bed aggradation replaced the previous degradation in the enlarged area (Habersack et al., 2003). Comparison of the main channel thalweg of October 2001 and June 2003 documented sedimentation up to 1 m (Fig. 27.5). In addition, substantial local erosion was observed at the conﬂuence of the main channel and sidearm. Flow velocity measurements and substrate analysis demonstrated the high variability in the main and side channel. Within the widened section, dynamic relocating processes now take place. Multiple river-bed structures and different habitats were formed. The dynamics of erosion and sedimentation processes are now extremely high. The monitoring of the restored river section showed that different aims can be reached concurrently: ﬂood protection, riverbed stabilisation, along with valuable new habitats for endangered animals and plants (Formann et al., in press). For the Upper Drau restoration measures, an increased density of juvenile ﬁsh species and a clear improvement of habitat quality has already been proven (Unfer et al., 2004; Muhar et al., this volume). Thus, the habitat quality for juvenile grayling was improved by initiating meso-habitat types such as shallow gravel-bars and coves: the 0+-densities are signiﬁcantly higher compared to regulated stretches of the Drau, where rip-rap is the dominating structure along the river banks (Unfer et al., 2004). Nevertheless, deﬁcits in the ecological situation remain. This includes the interruption of the longitudinal river continuum downstream by hydropower plants and surge effects (Muhar et al., this volume). Importantly, river-bed widening measures only stop bed degradation when sediment input takes place from upstream (sediment continuum) and enough gravel deposits (thickness of gravel layer) are available for the intended morphodynamics. Gravel mining in rivers is now generally forbidden in Austrian rivers and such a ban is a pre-requisite for successful river restoration by widening. Furthermore, instead of designing river restoration, the self-forming development of river morphology by side erosion has proven to be an ecologically valuable and sustainable measure. River-bed widening must be applied also in a river type-speciﬁc manner. For example, a meandering river type may not accept the widening due to unsuitable boundary conditions and therefore once again reduce the width (Habersack et al., 2004). 3.2.3.

Act on process relationships: sustainable management of bedload

In 1997, a master management plan was initiated by the State services and the regional water authorities of the Rhoˆne catchment. The plan indicated bedload as a

River restoration in the Alps and their surroundings

719

(a)

Flow direction

1st October 2001

10th June 2004

1st October 2001 (b)

Flow direction

Figure 27.5. Self-forming river widening and corresponding aggradation at the Drau River at KleblachLind: (a) aerial photographs, showing pre- and post-restoration conditions (Regional Government of Carinthia), (b) morphological changes due to river bed widening and self-forming side erosion; blue colour shows deep areas and white to grey colours indicate shallow areas; green is used for unsubmerged zones, (c) bed elevation changes, following restoration (Modiﬁed after Formann et al., 2004).

key element for river ecosystems, which must be managed in a sustainable manner. It promoted new approaches to bank erosion involving the development of erosion hazard mapping and the deﬁnition of corridors in which bank erosion could be allowed. This approach recognised that the deﬁcit in bed sediment was a consequence

H. Habersack, H. Pie´gay

720 10th June 2004

(b)

50 m

(c)

2001 570.00

2003 Nebenarm Sidearm

Zmin [a.s.l.]

569.00 568.00 567.00 566.00 565.00 564.00 563.00 562.00 35 34 33 32 31 30 29 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9

8

7

6

5

4

3

Longitudinal section (cross section numbers) Figure 27.5. (Continued)

of gravel mining activity, of channel clearing for ﬂood risk purposes, of long-term reduction in bedload supply from catchments due to vegetation encroachment following widespread afforestation (both planned and spontaneous), and of disconnected channel and hillslope sources. In this context, certain measures have been proposed to manage a dwindled resource. The nested approach (catchment scale/local scale) has been promoted in integrative management plans, which have been conducted at the catchment scale since 1992. Local scale sediment management is determined by processes at the catchment scale. In clearing their installations, the private managers of weirs, who traditionally removed the gravel from their reservoir to maintain water abstraction, now have to transfer the gravel immediately downstream of the weirs to preserve sediment transfer. In mountain areas, the RTM services had created numerous artiﬁcial areas for trapping and then systematically clearing gravel. They now also push the gravel downstream of these structures, particularly in reaches with a recognised deﬁcit and associated problems. In a reservoir of the upper Rhoˆne River (Seyssel)

River restoration in the Alps and their surroundings

721

which is silted by coarse sediment from the Les Usses River, gravel is captured by a pump and conveyed downstream of the dam by a pipe to maintain sediment continuity. On the Droˆme catchment, which has a general deﬁcit, local aggradation has been observed. Rather than removing the gravel from these areas, work was done within the active channel to create a temporary geometry (narrower channel with gravel levees) which can transport more gravel (higher shear stress in frequency and magnitude), locally reducing the ﬂood risk. Within projects funded by the European Community (Life Environment, Life Nature programs), scientists and managers are gaining preliminary experience in gravel reintroduction in the Ain and Droˆme Rivers. The feasibility of such restoration measures is also supported by previous management experiences such as the long-term sediment feeding of the Rhine River at Iffezheim and, after the construction of the hydropower plant Freudenau/Vienna, at the Austrian Danube downstream of Vienna (upto 300,000 m3 of gravel per year) (DonauConsult, 1998). In the Danube, the aim is to reduce this amount in future by a granulometric bed improvement (increase grain size; Habersack et al., 2007). The Ain River, in its downstream valley, underwent progressive erosion following a chain of dams built between 1933 and 1968. The downstream progression of the deﬁcit is estimated to average 500 m per year based on bar disappearance compared with historical aerial photos (Rollet et al., 2004). This is a major problem in term of ecological preservation, whereby the most valuable areas will be affected by this progression within the coming decade. In order to reduce this process and to restore the already disrupted reach, sediment reintroductions are being conducted. The potential bedload transport has been estimated based on hydraulic formulae, and the gravel sediment stored in the ﬂoodplain has been estimated from sediment cores and GIS calculations (Fig. 27.6). Based on these estimations, artiﬁcial reintroduction from ﬂoodplain storage would be a feasible strategy for several decades. Such an approach is efﬁcient in the sense that it will restore the bedload transport of the river (half of the potential annual bedload transport) and the associated habitats by coupling reintroduction with ﬂoodplain habitat restoration. The ﬁrst reintroduction took place in August 2005, using material derived from deepening and widening a former channel. Along the Droˆme River, a GIS procedure had been developed to identify the most efﬁcient potential sources of coarse sediment to supply the most critically degraded reaches. Criteria used include annual mean distance of sediment transport, lithological conditions, and connectivity between the channel network and the sediment source. From this map, 10 test sites were initially identiﬁed and 2 ultimately selected to test sediment reintroduction – 1 on a hillslope and 1 in the valley ﬂoor. In these two cases, the vegetation has been removed to reduce soil stability and to increase their sensitivity to surﬁcial and bank erosion (Lie´bault et al., 2001; Lie´bault, 2003). In some cases, it is not possible to improve discharge or sediment transfer. Here, restoration measures are proposed to improve the structure of the altered ecosystems, connecting the biotic communities with water. In the Rhoˆne basin, improving former channel restoration sites by deepening, widening, and reconnecting with the main channel has become a major challenge. The ﬁrst experimental site was restored in 1999, and the restoration of 80 former channels is underway upstream from Lyon.

H. Habersack, H. Pie´gay

722

Figure 27.6. Estimation of the sediment volumes stored in four ﬂoodplain areas (sites 1–4) of the Ain along the degraded reach Priay – Varambon (upper graph) and of the number of years needed by the rivers to transfer them downstream (lower graph) according to the mean annual bedload transport rate. A–D are different areas within the ﬂoodplain sites: (A) shrubby riparian areas, (B) shrubby riparian areas with local agricultural lands, (C) mainly agricultural areas, (D) other land uses (after Rollet et al., 2004).

The ﬁrst results demonstrate the ecological efﬁciency of these measures (Fig. 27.7, Amoros et al., 2005).

4.

Feedback on restoration measures

From the previous experience and debates in the scientiﬁc community, but also amongst managers, it is possible to open several discussions and identify challenges for the scientiﬁc community. An overview of impacts and possible restoration options is given in Table 27.5. 4.1. 4.1.1.

Challenges currently almost met Consider the dynamic aspects of rivers

According to Amoros and Petts (1993), the European approach has begun to focus on process-oriented restoration strategies. With feedback from previous restoration experiences, practitioners increasingly take into account process considerations, trying to work with nature as much as possible. The concept of self-restoration is increasingly applied, placing greater emphasis on the life span of recreated forms. The question of cost-beneﬁt of the actions is more frequently posed in a concrete

River restoration in the Alps and their surroundings

723

Figure 27.7. First feedback on the restoration program of the former channels of the Rhoˆne (example of the site of Pierre Be´nite where the former channel geometries were restored in 1999 and the minimum discharge was increased from 10 to 100 m3s1 in 2000). (a) Location of the three restored former channels along the by-passed section of the Rhoˆne, (b) total number of submerged vegetation species in each of the restored former channels, (c) number of submerged vegetation species only occurring in one of the restored former channels, (d) overbank sedimentation observed in the different restored former channels between 1999 and 2003, demonstrating the difference in physical evolution (and habitat conditions) of each of them (after Amoros et al., 2005).

724

Table 27.5.

Impacts and restoration options.

Restoration options

Impacts Riparian Upstream encroachment sediment in the active trapping bars

Note: 1 short-term measure 2 mid-term measure 3 long-term measure

X

Bank protections to prevent channel movement

Embankment/ Discharge impact on lowering in ﬂooding a by-passed channel

Surges/ discontinuum/ peak ﬂow lowering

Groundwater lowering and riparian dewatering

X X X X X

X

X X

X X X

X X X

X X X

X

X

X

X X X

X X X

X

X

X X X

X

X X

X

X

X

H. Habersack, H. Pie´gay

Land use change3 Reduce surge effects 1 Increase minimum ﬂow2 Improve river continuum2 Restore sediment3 continuum Open embankment1 Former channel restoration2 River bed widening2 Remove bank protection and allow side erosion2 Improve instream bed structure2 Raise the groundwater level (by ramping etc.)2 Remove dike to reactivate bank erosion2 Create erodible corridor to preserve dynamic banks1 Resupply artiﬁcial gravel2 Sustainably remove riparian vegetation2

In-channel works provoking degradation (mining, groins for navigation)

River restoration in the Alps and their surroundings

725

manner. Concerning natural processes governing river morphology, the following key restoration-relevant parameters can be mentioned: (i) land cover, (ii) hydrology, (iii) sediment regime and sediment transport, and (iv) morphodynamics (e.g., aggradation, degradation, meander migration, braiding). The example of the restoration of the Mur River in Austria at the border to Slovenia is meaningful in this context. Sediment transport and morphodynamic options and restrictions to river restoration were considered under constrained boundary conditions (Fig. 27.8). Given the chain of hydropower plants upstream, almost no gravel enters the study reach. Using a numerical sediment transport model, simulations of the future degradation of the reach over the next 60 years showed a clear ongoing trend (Hengl, 2001). Long sections of the Mur River have a gravel bed thickness of less than 0.5 m above the ﬁne material of the Tertiary. A bed breakthrough within the coming 60 years is a distinct possibility. In the short- to mid-term,

Suitability of areas for restoration measures distance to Tertiary less than 0.5 m Fixed usage of areas Mur-km 135

130

125

120

115

110

105

100

95

very well suited suited not suited

D

Measure type

DD

A A

C C

B B

C C

E

Bed level change [m]

0 -0.2 -0.4 -0.6 -0.8 -1 -1.2 135

130

125

120

115

110

105

100

300000 280000 260000 240000 220000 200000 180000 160000 140000 120000 100000 80000 60000 40000 20000 0 95

Gravel volume / 500 m length [m³]

Gravel volume/500 m length [m3] for a 100 m strip at Austrian side Historic degradation in m Simul. degrad. (60 years)

Mur-km Parameters for general positioning of measures

Figure 27.8. Sediment transport and morphodynamic options and restrictions to river restoration under constrained boundary conditions for the Mur River in Austria at the border to Slovenia; the blue line represents the gravel volume available for sediment feeding or self-forming side erosion during restoration; (A) river bed widening to about 200 m and artiﬁcial bedload supply, (B) self-forming side erosion and increase of river bed width, (C) initial river bed widening, followed by side erosion, (D) activation of gravel deposits without changing the effective bed width in combination with side arm reconnections, (E) alternative bed stabilisation by, e.g., ramps or local grain size increases; arrows show the length of the various measure types.

726

H. Habersack, H. Pie´gay

a change in reservoir operation aiming to improve sediment continuum is not realistic, so the option of activating gravel from the alluvium by river-bed widening and side erosion was tested. The numerical model showed that, for a period of 60 years, enough sediments are available to stop the existing degradation tendency. This strategy would also improve ecological functionality (Jungwirth et al., 2003). During these 60 years there would be time to develop and implement measures to improve the sediment input into this reach (e.g., by optimising the weirs and reservoir conditions to support gravel throughout during ﬂoods). As a sediment deﬁcit exists and degradation had already taken place, general river-bed widening and side erosion is not possible everywhere (Habersack and Schneider, 2000). First, if the gravel bed thickness above the tertiary ﬁne material is less than 0.5 m, bed break through remains a threat even if river-bed widening takes place, due to scouring processes (e.g., during bed form migration). If insufﬁcient gravel deposits are present along the river banks and inundation area, side erosion is limited. Hence, the relative position of the gravel layer compared to the river bed was evaluated to help identify suitable restoration areas. High priority has to be given to reaches with extreme degradation tendency. Secondly, land-use constrains possible restoration options, especially in places where settlements and infrastructure prohibit partial widening and side erosion. Finally, ﬁve types of measures were identiﬁed, whose numerical simulation showed overall positive effects (A: river bed widening to about 200 m and artiﬁcial bedload supply; B: self-forming side erosion and increase of river bed width; C: initial river bed widening, followed by side erosion; D: activation of gravel deposits without changing the effective bed width in combination with side arm reconnections; E: alternative bed stabilisation by, e.g., ramps or local grain size increases). A step-wise realisation of the measures was therefore suggested to optimise gravel availability and extend the lifetime of the measures, recognising that long-term gravel input from outside the reach is essential. 4.1.2.

Relativise the idea of a historical natural state as a goal to achieve

The historical natural state is still engrained as a reference in the minds of European river managers who designed the WFD. This reﬂects an ongoing search for reference reaches to evaluate the difference to the observed state, from which a score quantiﬁes ecosystem alteration. Whereas natural streams may be available upstream, this is rarely the case downstream. In relation to river restoration, this poses the question of adequate reference situations to deﬁne goals. Many restoration projects deﬁne conditions before human impacts as a vision for river restoration. In European projects, this often involves using maps from the 19th century as a reference, knowing that these do not fully reﬂect conditions before human impacts (e.g., humans considerably impacted some river systems before the 19th century, like deforestation on the slopes; 19th century river condition may reﬂect a different climatic period, such as the Little Ice Age), but are the only documents available. Individual maps, however, present a certain, time-speciﬁc view of the river under the boundary conditions that operated at that time and must be used under these boundary conditions. In Austria the so-called

River restoration in the Alps and their surroundings

727

Leitbild approach uses the reference state in two ways: (i) as a visionary goal, where process understanding rather than map description should dominate and (ii) as deﬁning the direction which the operational restoration measures target (regarding existing boundary conditions). Pragmatism must be applied in large projects such as the Danube, the Rhoˆne, or the Rhine, where managers are attempting to appropriately balance ecosystem improvement (best achievable) and the satisfaction of immediate human needs. A key goal is to develop participative management and integrate all actors. So-called ‘‘LIFE projects’’ are important in implementing river restoration measures.

4.1.3. Enlarge the scale to approach the problems (following the WFD and scientific advances) Currently, most managers have understood that a river cannot be managed like a state or a region without considering the natural boundaries and dynamics or the long-term adjustment. Managers in the WFD will work in hydrographic districts (river basins) and homogeneous water bodies. For example, bank erosion is considered more at a reach scale than at a local scale, and local bank protection without considering the overall plan is now prohibited in France. Gravel removal from channels can be permitted after establishing that the catchment has excess sediment, not a deﬁcit. In terms of future river basin management plans, the issue of scales in river restoration has been addressed in scientiﬁc terms, but not fully at the management level. This calls for a scientiﬁc advance towards methods of implementing scaling in practical management strategies (Habersack, 2000).

4.2. 4.2.1.

New challenges for river restoration Transform scaling to river restoration management

Theoretically, we know that it is essential to plan at the catchment scale and to nest the scales for successful river restoration, whereas in practise local to sectional projects continue to dominate. Furthermore, a practical scaling procedure remains to be implemented (Habersack, 2000). On the long term, various processes determine the status quo and patterns of rivers. A hierarchical dependency of small-scale processes on large-scale boundary conditions is given (Brierley and Fryirs, 2005). A successful, sustainable river restoration must include larger scales (especially the catchment-wide scale). A procedure for these scaling efforts is necessary. The WFD aims for river basin management, but no practical example pertaining to river restoration – that includes all relevant scales and deﬁnes a procedure for the necessary scaling steps – exists yet for the Alpine region. One possibility is the so-called River Scaling Concept (RSC), which suggests a two-step procedure (Habersack, 2000). In the ﬁrst, downscaling phase, the major boundary conditions and processes as well as patterns are analysed. In the second, upscaling phase, models are used to aggregate information in a way that allows detailed restoration measures to be suggested for each scale.

H. Habersack, H. Pie´gay

728 4.2.2.

Consider the physical factors to restore ecological functions

In Austria, water quality is no longer a problem, and mainly hydromorphological deﬁcits cause ecological difﬁculties. As shown in this paper, the main measures to improve the ecological status will involve improving the river continuum (biological), reducing surge effects and increasing minimum water ﬂow. This has already been shown for North American projects that target dam removal to re-establish the river continuum (e.g., Bednarek 2001; Hart and Poff, 2002). River morphological deﬁcits – especially concerning the sediment continuum – are known, but these are long-term developments and only become acute, e.g., in the ﬁsh population, when it is almost too late for ‘‘near natural’’ measures. In future, therefore, river morphological deﬁcits have to be considered as an essential part in restoring ecological functions. The required measures are cost intensive but unavoidable due to their central importance for sustainable river restoration. It would be necessary to extend the WFD to additionally include the monitoring of abiotic parameters so that negative developments can be recognised in time and timely action taken. Unfortunately, the decisive criteria for achieving good ecological status within the WFD are solely biological parameters, with hydromorphological elements supporting the biological elements. Even if mid-term trends towards catastrophic abiotic situations exist, as long as the biological metrics show a good ecological status, no actions need be taken. Sometimes, this may be too late (e.g., river bed breakthrough). 4.2.3.

Promote mitigating day-to-day management actions

Restoration often goes beyond spectacular actions. As indicated by Brierley and Fryirs (2005) from Australian case-studies, however, actions on ‘‘basket case reaches’’ (e.g., the worst condition sections) are very expensive, with the least likelihood of success and the risk of public disengagement. Sustainable management practices must also promote actions that are less ambitious but that should be widely implemented in a territory. One key issue in this domain is the improvement of measures performed by ﬂood managers acting on channel sediment storage (e.g., sand and gravel bars), ﬂoating wood and pioneer vegetation to prevent risk situations. Here, sustainable practices can be promoted, such as the sectorised riparian vegetation maintenance and wood removal as suggested by Pie´gay and Landon (1997) or by Boyer et al. (1998) to preserve ecological habitat provided by these structures. Moreover, when the ﬂood level is not critical (outside of settlements), vegetation and logjams contribute to increased roughness and related higher water levels and lower ﬂow velocities; this improves water retention upstream and reduces downstream ﬂood risk. In an era in which managers promote ﬂood management at catchment scales (e.g., the recent French law on Natural Risk published in July 2003), such properties of natural systems should be considered. Some river networks that have been strongly cleared for ﬂooding purposes probably create critical ﬂoods downstream due to decreased local roughness. Sustainable day-to-day vegetation management can be also promoted along reaches downstream of dams reducing peak ﬂow (i.e., Durance River in France). In the Alps and their surrounding areas, this context is so frequent that a general change

River restoration in the Alps and their surroundings

729

in practices would have major and immediate ecological beneﬁts. Whereas vegetation is traditionally removed periodically in a systematic band, creating a simpliﬁed corridor composed of a mature forest band and a wide gravel bar band, it is possible to complexify the vegetation removal temporally and spatially. This would recreate mechanically the successional stages resulting from bank erosion prior to dam construction (see Boyer and Pie´gay, 2003). Moreover, in catchments where gravel is becoming rare, sustainable practices must be promoted to preserve this material. Scientists have successfully promoted technical solutions for restoration (e.g., minimum discharge for optimum biological effects, best ﬂoodplain width for preserving bank erosion, best widening), but there is still a critical need to modify existing legislation (e.g., integrating the erodible corridor concept in French legislation and its consequences for mining authorisation). 4.2.4.

Link past and future in the restoration actions

Another challenge is to consider the past, not in terms of goals to achieve but as a source of information to understand the geomorphic and ecological trend, the adjustment conditions, the potential life span, recovery or diversity of recreated ecosystems – and then design the future. In many countries and for many rivers, the long-term morphodynamic channel evolution is not known. Beyond natural changes, anthropogenic inﬂuences on these parameters clearly must be analysed in detail in order to discuss the possibilities and restrictions of restoration measures. Process understanding is an essential component in order to analyse the historical development of river morphology and to predict future changes following river restoration. In relation to these issues, there is a need to promote prospective modelling and scenarios combining quantitative models with more qualitative information. This would provide a clear feedback of previous experiences to better target new actions, to better appreciate what works and what does not work in different geographical contexts. Establishing deﬁned monitoring programs and reference states is a challenge to evaluate the efﬁciency of measures. Often, decision-makers want to act rapidly despite a poorly described reference state. The sediment deﬁcit is very important here, and in many rivers it is impossible to reintroduce sediment. A strategy needs to be devised for such rivers. We must try to understand the direction in which rivers will develop without intervention. In some cases it will be necessary to accept that it is sometimes the only direction in which the river can go, and that no sustainable actions can modify such a trend. Scientists and managers know that technical answers have limits. Likewise, restoration measures also have limits and cannot be applied everywhere. This calls for identifying future problems associated with channel adjustment due to past human actions, the goal being to anticipate and adapt the local stakes and infrastructures to the new conditions. 4.2.5.

Promote a bridge between natural, technical, and social sciences

This debate needs to be opened – a difﬁcult task considering that spatial scales and methods differ between earth and biological sciences, technical, and social sciences. Do scientists promote restoration utopias? The scientiﬁc community must consider the feasibility of what it proposes. Up until what point can it promote a nested

H. Habersack, H. Pie´gay

730

perspective in terms of scaling? What is the cost-beneﬁt of the actions it proposes? What is the social acceptance of the suggested measures? If scientists strive to address a social need, they must consider this need properly. Over what period of time must the project design be valid? The overall community needs answers to these issues to better deﬁne technical questions and to abandon certain solutions. Scientists are experts who can inﬂuence the decision-making process. Today, the money comes from managers in the broadest sense: scientists may inﬂuence them in a particular direction that can promote research programs and yield results, but certain results may be utopian and ultimately inefﬁcient for society as a whole.

5.

Conclusions

This paper ﬁrst analyses the historical development of Alpine rivers with respect to hydromorphological and ecological deﬁcits. Based on a review of existing restoration projects in the Alps and their surrounding areas, currently solved and remaining challenges are discussed. The following conclusions are drawn: (i) in the Alps, every larger river system is anthropogenically inﬂuenced, (ii) many rivers have already reached or will soon reach a critical state of morphodynamic development (e.g., river bed breakthrough), where ‘‘natural’’ river restoration will be almost impossible, (iii) sediment transport and river morphodynamics play a central role in river restoration and need to be incorporated, including ﬂoodplain restoration, (iv) a link must be drawn between the past and the future with respect to restoration actions, (vi) implementing the European Water Framework Directive will promote river restoration, the goal being to reach good ecological status of running waters by 2015, (vi) beside ecological parameters (e.g., composition, abundance, and age structure of the ﬁsh fauna), hydromorphological variables should be also included in the monitoring programs of the WFD to evaluate the development of rivers and to promptly react to critical trends, (vii) a scale-oriented approach has to be developed to practically implement river restoration, including scaling (from the catchment-wide scale to the point scale), (viii) future restoration measures should involve major individual measures but also day-to-day management actions, and (ix) a bridge between natural, technical, and social sciences is crucial for successful river restoration, taking a cross disciplinary approach ranging from river engineering, landscape, and areal planning to biology.

Acknowledgements The authors wish to thank the Austrian Ministry of Agriculture, Forestry, Environment and Water Management, along with Regional Governments and local authorities, the European Community (Life Nature ‘‘Restoration of the wetland and riparian area at the Upper Drau River’’ 1999–2003), the Water Agency RMC, the region Rhoˆne-Alpes and the local managers for ﬁnancial support of various research projects related to river restoration (mainly Life Environment ‘‘Forests for Water’’

River restoration in the Alps and their surroundings

731

2003–2007 and Life Nature ‘‘Conservation des habitats de la basse valle´e de l’Ain’’ 2003–2007). E. Formann and St. Schober supported the paper by ﬁeldwork and helped prepare certain ﬁgures.

References Amoros, C., 2001. The concept of habitat diversity between and within ecosystems applied to river sidearm restoration. Environ. Manage. 28 (6), 805–817. Amoros, C., Elger, A., Dufour, S., et al., 2005. Flood scouring and groundwater supply in side-channel rehabilitation of the Rhoˆne River. Archiv fu¨r Hydrobiologie, France 155, 147–167. Amoros, C. and Petts, G.E. (Eds), 1993. Hydrosyste`mes ﬂuviaux. Coll. d’e´cologie – 24, Masson, Paris, 300pp. published in English in 1994. Amoros, C., Richardot-Coulet, M., Pautou, G., 1982. Les ‘‘ensembles fonctionnels ‘‘: des entite´s e´cologiques qui traduisent l’e´volution de l’hydrosyste`me en inte´grant la ge´omorphologie et l’anthropisation (exemple du Haut-Rhoˆne franc- ais). Rev. Ge´ogr. Lyon 57 (1), 49–62. Bednarek, A.T., 2001. Undamming rivers: a review of the ecological impacts of dam removal. Environ. Manage. 27, 803–814. Binder, W., Ju¨rging, P., Karl, J., 1983. Natural river engineering – characteristics and limitations. Garten Landschaft (Landsc. Archit. Plann.) 2, 91–98. Bornette, G., Heiler, G., 1994. Environmental and biological responses of former channels to river incision: a diachronic study on the upper Rhoˆne river. Regul. Rivers Res. Manage. 9, 79–92. Bourdin, L., 2004. Les rivie`res en tresses sur le basin Rhoˆne Me´diterrane´e Corse. Bilan et perspectives de gestion. Me´m.de ﬁn d’e´tude. Master Gestion de l’Eau, ENGREF, CNRS. 60pp. Boyer, M., Pie´gay, H., 2003. Re´habilitation, restauration et entretien des ripisylves. In: Pie´gay, H., Pautou, G., and Rufﬁnoni, C. (Eds), Les foreˆts riveraines des cours d’eau: e´cologie, fonctions, gestion. Institut pour le De´veloppement Forestier, Paris, pp. 390–413. Boyer, M., Pie´gay, H., Rufﬁnoni, C., et al., 1998. Guide technique SDAGE no. 1 – Vol. 1: La gestion des boisements de rivie`re: dynamique et fonctions de la ripisylve,’’ 49pp, and Vol. 2: La gestion des boisements de rivie`re – De´ﬁnition et sectorisation des objectifs de gestion et des niveaux d’entretien,’’ Agence de l’Eau Rhoˆne Me´diterrane´e Corse, 56pp+annexes. Braga, G., Gervasoni, S., 1989. Evolution of the Po river: an example of the application of historic maps. In: Petts, G.E. (Ed.), Historical Changes of Large Alluvial Rivers: Western Europe. Wiley, Chichester, pp. 127–143. Bravard, J.P., Amoros, C., Pautou, G., 1986. Impact of civil engineering works on the successions of communities in a ﬂuvial system. A methodological and predictive approach applied to a section of the Upper Rhone river, France. Oikos 47 (1), 92–111. Bravard, J.P., Amoros, C., Pautou, G., et al., 1997. River incision in South-east France: morphological phenomena and ecological effects. Regul. Rivers Res. Manage. 13, 1–16. Bravard, J.P., Kondolf, G.M., Pie´gay, H., 1999a. Environmental and societal effects of river incision and remedial strategies. In: Simon, A. and Darby, S. (Eds), Incised River Channels. Wiley, Chichester, pp. 303–341. Bravard, J.P., Landon, N., Peiry, J.L., Pie´gay, H., 1999b. Principles of engineering geomorphology for managing river erosion and bedload transport, examples from French rivers. Geomorphology 31, 291–311. Bravard, J.P., Peiry, J.L., 1993. La disparition du tressage ﬂuvial dans les Alpes franc- aises sous l’effet de l’ame´nagement de cours d’eau (19-20e`me sie`cle). Zeitschrift fu¨r Geomorphologie, Suppl.-Bd. 88, 67–79. Brierley, G.J., Fryirs, K.A., 2005. Geomorphology and River Management: Applications of the River Styles Framework. Blackwell Publications, Oxford, UK, 398pp. Brookes, A., 1987. Restoring the sinuosity of artiﬁcially straightened stream channels. Environ. Geol. Water Sci. 10 (1), 33–41.

732

H. Habersack, H. Pie´gay

Brookes, A., 1990. Restoration and enhancement of engineered river channels: some European experiences. Regul. Rivers Res. Manage. 6, 45–56. BUWAL, 2002. Umwelt-Bericht Bd 1. Politik und Perspektiven des BUWAL. Bundesamt fu¨r Umwelt, Wald und Landschaft. Citterio, A., Rollet, A.J., Pie´gay, H., 2002. Synthe`se des connaissances ge´omorphologiques pour e´valuer la diversite´ physique et l’e´tat de de´gradation des bras morts du Doubs a` l’aval de Dole. Unpublished report, DIREN Franche-Comte´, 39pp. Dister, E., 1992. La maıˆ trise des crues par la renaturation des plaines alluviales du Rhin supe´rieur. Espaces naturels rhe´nans Bull. Soc. Ind. de Mulh. 824-1, 73–82. DonauConsult, Zottl & Erber und A. Oberhofer, 1998. Donauausbau o¨stlich von Wien. Flussbauliches Gesamtprojekt; Vorprojekt, Strom-km 1910 bis 1895. Downs, P.W., Gregory, K.J., 2004. River channel management: towards sustainable catchment hydrosystems, ISBN: 0340759690, Arnold Publication, 256pp. Edouard, J.L., Vivian, H., 1984. Une hydrologie naturelle dans les Alpes du Nord? Les nouveaux parame`tres de l’hydrologie alpine: les ame´nagements hydroe´lectriques. Rev. Ge´ogr. Alpine 72 (2–3), 165–188. European Parliament and Council of Europe, 2000. Directive 2000/60/EC of the European Parliament and of the Council of 23 October 2000 establishing a framework for Community action in the ﬁeld of water policy. Ofﬁcial Journal L 327, 22/12/2000, 73pp. Formann, E., Habersack, H. M., Schober, St. (2007). Morphodynamic river processes and techniques for assessment of channel evolution in Alpine gravel bed rivers. Geomorphology (in press), doi:10.1016/ j.geomorph.2006.10.029. Formann, E., Schober, St., Habersack, H.M., 2004. Monitoring and modelling of river widening processes in an alpine river. Ninth International Symposium on River Sedimentation, October 18–21, 2004, Vol. 3, pp. 1627–1634, Yichang, China. Frissell, C.A., Liss, W.J., Warren, C.E., Hurley, M.D., 1986. A hierarchical framework for stream habitat classiﬁcation: viewing streams in a watershed context. Environ. Manage. 10, 199–214. Garbrecht, G., 1985. Wasser. Rowohlt TB-Verlag. Reinbek, Germany. Gilvear, D.J., 1999. Fluvial geomorphology and river engineering: future roles utilizing a ﬂuvial hydrosystems framework. Geomorphology 31, 229–245. Girel, J., Garguet-Duport, B., Pautou, G., 1997. Landscape structure and historical processes along diked European valleys: a case study of the Arc/Ise`re conﬂuence (Savoie, France). Environ. Manage. 21 (6), 891–907. Glitz, D., 1983. Artiﬁcial channels – the ‘‘ox-bow’’ lakes at tomorrow: the restoration of the course of the Wandse in Hambourg-Rahlstedt. Garten Landschaft (Landsc. Archit. Plann.) 2, 109–111. Habersack, H., Liedermann, M. and Tritthart, M., 2007. Restoring large rivers – the integrated Danube river project. International Conference Ecohydraulics 2007, 18.–23.02.07, New Zealand. Habersack, H.-M., 2000. The river scaling concept (RSC): a basis for ecological assessments. J. Hydrobiol. 422/423, S. 49–S. 60. Habersack, H.-M., Hauer, Ch., Edin, S., et al., 2004. Ecologically oriented hydraulic engineering at an Austrian lowland river: inﬂuences of artiﬁcial structures on ﬂood processes and the morphological development. In: Proceedings of the 10th international Congress – INTERPREAVENT: 24. – 27.05.2004, Riva del Garda; Trient, Italy. Habersack, H.-M., Koch, K., Nachtnebel, H.-P., 2000. Flussaufweitungen in O¨sterreich – Entwicklung, Stand und Ausblick. O¨sterreichische Wasser und Abfallwirtschaft, O¨WAV 52, 143–153. Habersack, H.-M., Laronne, J.B., 2001. Bedload texture in an Alpine gravel bed river. Water Resour. Res. 37 (12), 3359–3370. Habersack, H.-M., Nachtnebel, H.-P., 1995. Short-term effects of local river restoration on morphology, ﬂow ﬁeld, substrate and biota. Regul. Rivers Res. Manage. 10, 291–301. Habersack, H.-M., Nachtnebel, H.-P., 1998. Planung und Konzeption ﬂussbaulicher MaXnahmen zur Sohlsicherung und Verbesserung der gewa¨ssermorphologischen Strukturen. O¨sterreichische Wasserund Abfallwirtschaft, Heft 1/2, Jg. 50, S. 40–S. 48. Habersack, H.-M., Schneider, J., 2000. Ableitung und Analyse ﬂussmorphologisch relevanter Parameter von historischen Karten. Wasser & Boden, 52. Jg., 6/2000, S. 55–S. 59.

River restoration in the Alps and their surroundings

733

Habersack, H.-M., Schober, S.T., Formann, E., et al., 2003. Flussmorphologisches Monitoring im Rahmen des LIFE-Projektes ‘‘Obere Drau’’. In: (Hrsg.): BMLFUW: 20. Flussbautagung LIFESYMPOSIUM: Gewa¨sserbetreuung und die EU-Wasserrahmenrichtlinie—Umsetzung am Beispiel von LIFE Projekten, Sept. 2003, Spittal/Drau, S. 15–45, Wien. Hart, D.D., Poff, N.L., 2002. A special section on dam removal and river restoration. Bioscience 52, 653–655. Hengl, M., 2001. Numerical simulation of bedload transport at the Mur river between Spielfeld and Radkersburg, ﬁnal report for the Austrian Ministry of Agriculture, Forestry, Environment and Water Management and Styrian water authority, unpublished. 120pp. Hengl, M., 2004. Processes in river morphology and extreme ﬂoods, presentation at the conference FloodRisk of the Austrian Ministry of Agriculture, Forestry, Environment and Water Management, Vienna. Henry, C.P., Amoros, C., 1995a. Restoration ecology of riverine wetlands. I. A scientiﬁc base. Environ. Manage. 19 (6), 891–902. Henry, C.P., Amoros, C., 1995b. Restoration ecology of riverine wetlands. II. An example in a former channel of the Rhone river. Environ. Manage. 19 (6), 903–913. Hunzinger, L.M., 1998. Flussaufweitungen – Morphologie, Geschiebehaushalt und Grundsa¨tze zur Bemessung. Vol. 159, VAW Mitteilungen, Versuchsanstalt fu¨r Wasserbau, Hydrologie und Glaziologie der ETH Zu¨rich, Zu¨rich. Iversen, T.M., Kronvang, B., Madsen, B.L., et al., 1993. Re-establishment of Danish streams: restoration and maintenance measures. Aquat. Conserv. Mar. Freshw. Ecosyst. 3, 73–92. Ja¨ggi, M., Zarn, B., 1999. Stream channel restoration and erosion control for incised channels in Alpine environments. In: Darby, S. and Simon, A. (Eds), Incised River Channels. Wiley, Chichester, UK, pp. 343–370. Ja¨ggi, M.N.R., 1989. Channel engineering and erosion control. In: Gore, J. and Petts, G. (Eds), Alternative in Regulated River Management. CRC Press, Boca Raton, FL. Jungwirth, M., 1986. Flussbau und Fischerei. Quantitative Untersuchungen u¨ber die Auswirkungen unterschiedlicher Flussregulierungen auf Fischbesta¨nde. Wien. Mitteilungen 70, 129. Jungwirth, M., Haidvogl, G., Moog, O., et al., 2003. Angewandte Fischo¨kologie an FlieXgewa¨ssern. UTB, Facultas Universita¨tsverlag, Viennapp, 547pp. Jungwirth, M., Moog, O., Muhar, S., 1993. Effects of river bed restructuring on ﬁsh and benthos of a ﬁfth order stream, Melk, Austria. Regul. Rivers Res. Manage. 8, 195–204. Klein, J.P., Maire, G., Exinger, F., et al., 1994. The restoration of former channels in the Rhine alluvial forest: the example of the Offendorf nature reserve (Alsace, France). Water Sci. Tech. 29 (3), 301–305. Kondolf, G.M., Downs, P.W., 1996. Catchment approach to planning channel restoration. In: Brookes, A. and Shields, F.D. Jr. (Eds), River Channel Restoration: Guiding Principles for Sustainable Projects. Wiley & Sons Ltd., Chichester, UK, pp. 129–148. Kondolf, M.G., Pie´gay, H., Landon, N., 2007. Changes in the riparian zone of the lower Eygues river, France, since 1830. Landsc. Ecol. 22 (3), 367–384. Lebensministerium, 2005. Status of Austrian surface waters, evaluation results according to the EU Water Framework Directive. Lie´bault, F., 2003. Les rivie`res torrentielles des montagnes droˆmoises: e´volution contemporaine et fonctionnement ge´omorphologique actuel (massifs du Diois et des Baronnies). PhD thesis, Universite´ Lumie`re Lyon 2, Lyon. Lie´bault, F., Cle´ment P. and Pie´gay, H., 2001. Analyse ge´omorphologique de la recharge se´dimentaire des bassins versants de la Droˆme, de l’Eygues et du Roubion. Final report, Ofﬁce National des Foreˆts, De´partement de la Droˆme (DDA) & Agence de l’Eau RMC, 2 vol.+maps. Lie´bault, F., Pie´gay, H., 2002. Causes of 20th century channel narrowing in mountain and piedmont rivers and streams of Southeastern France. Earth Surf. Process. Landf. 27, 425–444. Martinet, F., Dubost, M., 1992. Die letzten naturnahen Alpenﬂu¨sse, Vol. 11, kleine Schriften, Internationale Alpenschutzkommission CIPRA, Vaduz. Mo¨ller, H.M., Wefers, K., 1983. The restoration of cut-off river channels as illustrated by the lower reaches of the Kru¨ckau in the District of Pinneberg. Garten Landschaft (Landsc. Archit. Plann.) 2, 107–108. Muhar, S., Jungwirth, M., Unfer, G., et al., in press. Restoring riverine landscapes – success and deﬁcits in the context of ecological integrity, this volume.

734

H. Habersack, H. Pie´gay

Muhar, S., Kainz, M., Kaufmann, M., Schwarz, M., 1998. Erhebung und Bilanzierung ﬂusstypspeziﬁsch erhaltener FlieXgewa¨sserabschnitte in O¨sterreich. O¨sterreichische Wasser- und Abfallwirtschaft, O¨WAV 50 (5–6), 119–127. Peiry, J.L., 1987. Dynamique ﬂuviale historique de l’Arve dans le bassin de Cluses (Haute-Savoie). 112e`me Congre`s national des Socie´te´s savantes, Lyon. Peiry, J.-L., Marnezy, A., 2000. Les barrages et re´servoirs hydroe´lectriques des Alpes Franc- aises et leurs impacts sur les cours d’eau. In: Bravard, J.-P. (Ed.), Les re´gions franc- aises face aux extreˆmes hydrologiques; gestion des exce`s et de la pe´nurie. Editions SEDES, Paris, pp. 190–209. Peiry, J.L., Salvador, P.G., Nouguier, F., 1994. L’incision des rivie`res des Alpes du Nord : e´tat de la question. Rev. Ge´ogr. Lyon 69 (1), 47–56. Petts, G.E., 1989. Historical analysis of ﬂuvial hydrosystems. In: Petts, G.E. (Ed.), Historical Change of Large Alluvial Rivers, Western Europe. Chichester, Great Britain, pp. 1–19. Pie´gay, H., Cuaz, M., Javelle, E., Mandier, P., 1997. A new approach to bank erosion management: the case of the Galaure river, France. Regul. Rivers Res. Manage. 13, 433–448. Pie´gay, H., Darby, S.A., Mosselmann, E., Surian, N., 2005. The erodible corridor concept: applicability and limitations for river management. River Res. Appl. 21, 773–789. Pie´gay, H., Landon, N., 1997. Promoting an ecological management of riparian forests on the Droˆme River, France. Aquat. Conserv. Mar. Freshw. Ecosyst. 7, 287–304. Pie´gay, H., Salvador, P.G., 1997. Contemporary ﬂoodplain forest evolution along the middle Ubaye river. Global Ecol. Biogeogr. Lett. 6/5, 397–406. Pie´gay, H., Schumm, S.A., 2003. System approach in ﬂuvial geomorphology. In: Kondolf, M.G. and Pie´gay, H. (Eds), Tools in Fluvial Geomorphology. Wiley, Chichester, UK, pp. 105–134. Pinay, G., Haycock, N.E., Rufﬁnoni, C., Holmes, R.M., 1994. The role of denitriﬁcation in nitrogen removal in river corridors. In: Mitsch, W.J. (Ed.), Global Wetlands: Old World and New. Elsevier Science, Amsterdam, pp. 107–116. Poinsart, D., Salvador, P.G., 1993. Histoire de l’endiguement du Rhoˆne a` l’aval de Lyon (19th c.). In: Piquet, F. (Ed.), Le ﬂeuve et ses me´tamorphoses. Didier Erudition, Lyon. France, pp. 299–313. Qiu, Z., Prato, T., 1998. Economic evaluation of riparian buffers in an agricultural watershed. J. Am. Water Resour. Assoc. 34 (4), 877–890. Reckendorfer, W., Schmalfuss, R., Baumgartner, C., et al., 2005. The Integrated River Engineering Project for the free-ﬂowing Danube in the Austrian Alluvial Zone National Park: framework conditions, decision process and solutions. In: Buijse, A.D. (Ed.), Lowland River Rehabilitation 2003, Sept. 29–Oct. 2, Wageningen, Netherlands; Archiv fu¨r Hydrobiologie, Large Rivers Supplement. Rohde, S., 2004. River restoration: Potential and limitations to re-establish riparian landscapes. Assessment and Planning, dissertation at the ETH Zu¨rich, Diss. ETH No. 15496, 127pp. Rollet, A.J., Lejot, J., Pie´gay, H., 2004. Basse Valle´e de l’Ain : expertise hydroge´omorphologique en vue du diagnostic fonctionnel des habitats, de la restauration du transit se´dimentaire et des loˆnes. Rapport ﬁnal. Programme Life-Nature Basse valle´e de l’Ain, SIVU du bassin de la basse valle´e de l’Ain, Communaute´ Europe´enne, Agence de l’Eau RMC, 40pp. Roux, A.L., Bravard, J.P., Amoros, C., Pautou, G., 1989. Ecological changes of the French upper Rhoˆne River since 1750. In: Petts, G.E. (Ed.), Historical Change of Large Alluvial Rivers, Western Europe, Great Britain, pp. 323–350. Rutherfurd, I., Ladson, A., Tilleard, J., et al., 1998. Research and development needs for river restoration in Australia. LWRRDC Occasional Paper 15/98. 91pp. ISBN 0642-26728-6. Schiemer, F., Baumgartner, C., Tockner, K., 1999. Restoration of ﬂoodplain rivers: the ‘‘Danube restoration project’’. Regul. Rivers Res. Manage. 15, 231–244. Sear, D.A., 1994. River restoration and geomorphology. Aquat. Conserv. Mar. Freshw. Ecosyst. 4, 169–177. Stephan, U., Hengl., M., Schaipp, B., 2003. River restoration considering geomorphological boundary conditions. In: Ganoulis, J. and Prinos, P. (Eds), Proceedings of the XXX IAHR Congress (Water engineering and research in a learning society: modern developments and traditional concepts), Thessaloniki, Greece Theme C: Reservoir Sedimentation, ISBN 960-243-602-6, 445–452. 458–464. Surian, N., Rinaldi, M., 2003. Morphological response to river engineering and management in alluvial channels in Italy. Geomorphology 50, 307–326.

River restoration in the Alps and their surroundings

735

Tockner, K., Schiemer, F., Ward, J.V., 1998. Conservation by restoration: the management concept for a river-ﬂoodplain system on the Danube River in Austria. Aquat. Conserv. Mar. Freshw. Ecosyst. 8, 71–86. Tricart, J., Bravard, J.P., 1991. Le cours pe´rialpin du Rhin, du Rhoˆne et du Danube: ame´nagement ﬂuvial et de´rives de l’environnement. Ann. Ge´ogr. 561–562, 668–713. Unfer, G., Schmutz, St., Wiesner, Ch., et al., 2004. The effects of hydropeaking on the success of riverrestoration measures within the LIFE-project ‘‘Auenverbund Obere Drau’’. In: de Jalon, D.G., Vizcaino Martinez, P. (Eds), Fifth International Symposium on Ecohydraulics, 12.09.2004-17.09.2004, Madrid; Proceedings of the Fifth International Conference on Ecohydraulics – Aquatic Habitats: Analysis and Restoration, 1, Madrid; ISBN 90-805649-7-4, 741–746. Vischer, D., 1986. Schweizerische Flusskorrektionen im 18. und 19. Jahrhundert, Vol. 84, VAW Mitteilungen, Zu¨rich. Vischer, D., 2003. Die Geschichte des Hochwasserschutzes in der Schweiz. Von den Anfa¨ngen bis ins 19. Jahrhundert. Berichte des BWG, Serie Wasser, Bundesamt fu¨r Wasser und Geologie (BWG), 208. Wasson, J.G., Malavoi, J.R., Maridet, L., et al., 1998. Impacts e´cologiques de la chenalisation des rivie`res. Etudes, gestion des milieux aquatiques 14, Lyon, 157pp.

Discussion by G. Williams The aim is self-forming river restoration. However, in the Austrian rivers we saw, the river restoration reaches were very short and at irregular intervals down the river. These short widenings increase the inconsistency of sediment movement (in concert with the sudden changes in channel form), adding local aggradation/degradation steps. The river will then actively re-work these areas, and this, along with the natural vegetational changes, will cause ongoing changes to the widened area. Thus, for the ecological values of the wider reaches to be maintained, ongoing interventions will be required. In what way, then, are these short widenings self-forming or self-maintaining? Reply by the authors Gary Williams raises an interesting question. The feedback we have in terms of riverbed widening is relatively short (up to 15 years) and this allows us to only partially assess the adjustment of the riparian vegetation. The question of self-maintaining channel widenings in discontinuous short reaches within a long regulated reach is an important issue and a challenging research question that is best answered by longterm monitoring programs. At this stage, we can only offer the optimistic arguments for self-maintenance of these new features within the context of Austria. We also present and discuss a pessimistic point of view. In this context, self-forming river restoration means that only initial measures are realised through human intervention and that further development of the morphology is accomplished by the river itself. For the reach described in the paper (see Fig. 27.5) the initial measures were the removal of rip-rap and the creation of a side channel (Fig. 27.9 below). Immediately afterwards, intensive bank erosion occurred. A signiﬁcant increase in channel width combined with channel bed aggradation was observed (Fig. 27.10

River restoration in the Alps and their surroundings

735

Tockner, K., Schiemer, F., Ward, J.V., 1998. Conservation by restoration: the management concept for a river-ﬂoodplain system on the Danube River in Austria. Aquat. Conserv. Mar. Freshw. Ecosyst. 8, 71–86. Tricart, J., Bravard, J.P., 1991. Le cours pe´rialpin du Rhin, du Rhoˆne et du Danube: ame´nagement ﬂuvial et de´rives de l’environnement. Ann. Ge´ogr. 561–562, 668–713. Unfer, G., Schmutz, St., Wiesner, Ch., et al., 2004. The effects of hydropeaking on the success of riverrestoration measures within the LIFE-project ‘‘Auenverbund Obere Drau’’. In: de Jalon, D.G., Vizcaino Martinez, P. (Eds), Fifth International Symposium on Ecohydraulics, 12.09.2004-17.09.2004, Madrid; Proceedings of the Fifth International Conference on Ecohydraulics – Aquatic Habitats: Analysis and Restoration, 1, Madrid; ISBN 90-805649-7-4, 741–746. Vischer, D., 1986. Schweizerische Flusskorrektionen im 18. und 19. Jahrhundert, Vol. 84, VAW Mitteilungen, Zu¨rich. Vischer, D., 2003. Die Geschichte des Hochwasserschutzes in der Schweiz. Von den Anfa¨ngen bis ins 19. Jahrhundert. Berichte des BWG, Serie Wasser, Bundesamt fu¨r Wasser und Geologie (BWG), 208. Wasson, J.G., Malavoi, J.R., Maridet, L., et al., 1998. Impacts e´cologiques de la chenalisation des rivie`res. Etudes, gestion des milieux aquatiques 14, Lyon, 157pp.

Discussion by G. Williams The aim is self-forming river restoration. However, in the Austrian rivers we saw, the river restoration reaches were very short and at irregular intervals down the river. These short widenings increase the inconsistency of sediment movement (in concert with the sudden changes in channel form), adding local aggradation/degradation steps. The river will then actively re-work these areas, and this, along with the natural vegetational changes, will cause ongoing changes to the widened area. Thus, for the ecological values of the wider reaches to be maintained, ongoing interventions will be required. In what way, then, are these short widenings self-forming or self-maintaining? Reply by the authors Gary Williams raises an interesting question. The feedback we have in terms of riverbed widening is relatively short (up to 15 years) and this allows us to only partially assess the adjustment of the riparian vegetation. The question of self-maintaining channel widenings in discontinuous short reaches within a long regulated reach is an important issue and a challenging research question that is best answered by longterm monitoring programs. At this stage, we can only offer the optimistic arguments for self-maintenance of these new features within the context of Austria. We also present and discuss a pessimistic point of view. In this context, self-forming river restoration means that only initial measures are realised through human intervention and that further development of the morphology is accomplished by the river itself. For the reach described in the paper (see Fig. 27.5) the initial measures were the removal of rip-rap and the creation of a side channel (Fig. 27.9 below). Immediately afterwards, intensive bank erosion occurred. A signiﬁcant increase in channel width combined with channel bed aggradation was observed (Fig. 27.10

River restoration in the Alps and their surroundings

735

Tockner, K., Schiemer, F., Ward, J.V., 1998. Conservation by restoration: the management concept for a river-ﬂoodplain system on the Danube River in Austria. Aquat. Conserv. Mar. Freshw. Ecosyst. 8, 71–86. Tricart, J., Bravard, J.P., 1991. Le cours pe´rialpin du Rhin, du Rhoˆne et du Danube: ame´nagement ﬂuvial et de´rives de l’environnement. Ann. Ge´ogr. 561–562, 668–713. Unfer, G., Schmutz, St., Wiesner, Ch., et al., 2004. The effects of hydropeaking on the success of riverrestoration measures within the LIFE-project ‘‘Auenverbund Obere Drau’’. In: de Jalon, D.G., Vizcaino Martinez, P. (Eds), Fifth International Symposium on Ecohydraulics, 12.09.2004-17.09.2004, Madrid; Proceedings of the Fifth International Conference on Ecohydraulics – Aquatic Habitats: Analysis and Restoration, 1, Madrid; ISBN 90-805649-7-4, 741–746. Vischer, D., 1986. Schweizerische Flusskorrektionen im 18. und 19. Jahrhundert, Vol. 84, VAW Mitteilungen, Zu¨rich. Vischer, D., 2003. Die Geschichte des Hochwasserschutzes in der Schweiz. Von den Anfa¨ngen bis ins 19. Jahrhundert. Berichte des BWG, Serie Wasser, Bundesamt fu¨r Wasser und Geologie (BWG), 208. Wasson, J.G., Malavoi, J.R., Maridet, L., et al., 1998. Impacts e´cologiques de la chenalisation des rivie`res. Etudes, gestion des milieux aquatiques 14, Lyon, 157pp.

Discussion by G. Williams The aim is self-forming river restoration. However, in the Austrian rivers we saw, the river restoration reaches were very short and at irregular intervals down the river. These short widenings increase the inconsistency of sediment movement (in concert with the sudden changes in channel form), adding local aggradation/degradation steps. The river will then actively re-work these areas, and this, along with the natural vegetational changes, will cause ongoing changes to the widened area. Thus, for the ecological values of the wider reaches to be maintained, ongoing interventions will be required. In what way, then, are these short widenings self-forming or self-maintaining? Reply by the authors Gary Williams raises an interesting question. The feedback we have in terms of riverbed widening is relatively short (up to 15 years) and this allows us to only partially assess the adjustment of the riparian vegetation. The question of self-maintaining channel widenings in discontinuous short reaches within a long regulated reach is an important issue and a challenging research question that is best answered by longterm monitoring programs. At this stage, we can only offer the optimistic arguments for self-maintenance of these new features within the context of Austria. We also present and discuss a pessimistic point of view. In this context, self-forming river restoration means that only initial measures are realised through human intervention and that further development of the morphology is accomplished by the river itself. For the reach described in the paper (see Fig. 27.5) the initial measures were the removal of rip-rap and the creation of a side channel (Fig. 27.9 below). Immediately afterwards, intensive bank erosion occurred. A signiﬁcant increase in channel width combined with channel bed aggradation was observed (Fig. 27.10

H. Habersack, H. Pie´gay

736 6th June 2002

Initial side channel

Figure 27.9. Initial side channel as the beginning of a self-forming river restoration (based on Formann et al., 2004).

deposition erosion 45,5 m 25,0 m 4 m²

WSP 10 m²

49 m²

2,7 m

Figure 27.10. Self-forming river restoration at Kleblach (Based on Habersack et al., 2007).

below). Over time, the river actively re-worked the area and vegetational changes occurred (Formann et al., in press). It was shown that the ecological status was signiﬁcantly improved as a result of these self-formed morphological changes (Muhar et al., this volume), and no ongoing interventions have been required. In fact, due to the minimisation of the bed degradation – one of the major aims of the measures – the maintenance costs for repairing bank protection measures etc. have been signiﬁcantly reduced. To date, the total length of the restored sections in the Drau River is 10 km of an overall length of 70 km. Further sections will be restored over the coming years within a new EU Life project, with the ultimate goal covering the entire length. The length of the individual restoration reaches is a crucial point, because sections that are too short do not create a self-regulated morphological unit due to inﬂuence from upstream and downstream boundary conditions. In order to deﬁne the minimum length of a self-maintaining partial braiding, in addition to numerical models (Habersack et al., 2007), an engineering approach (Hunzinger, 1998) was used during the planning stages: Lw ¼

BA BK ½2:21 2:81 lnð1 F Þ 2

(27.1)

River restoration in the Alps and their surroundings F ¼ 0:21e7:1=ðbþ3:5Þ

737 (27.2)

BA BK

(27.3)

Lmin ¼ 2Lw

(27.4)

b¼

where Lw is the length of unprotected banks after self-forming bank erosion and river bed widening, BA the river bed width within the widening, BK the width of the regulated channel, F a parameter to calculate the ﬂow spread, b the ratio of widened over regulated width, and Lmin the minimum length of the river-bed widening/ restoration reach shown in Fig. 27.9. Equation (27.4) calculates a minimum length. According to physical model studies and practical experiences, 4 Lw is the most realistic value. For the Drau River, the minimum length is thus about 600 m. The main restoration reaches are longer, e.g., the reach shown in Fig. 27.10 has a length of more than 2 km. Following G. Williams’ questions, we also present some pessimistic arguments for self-maintenance. These arguments mainly apply if the bedload supply is limited and the peak ﬂow reduced, which is not the case for the Drau River. If bedload supply, peak ﬂow, and also the mean water discharge during summer are signiﬁcantly reduced, we can expect that riparian vegetation will encroach on the bars and recreate local geomorphic conditions similar to those observed in the embanked reaches. Besides permanent wetting, the maintenance of gravel bars is linked to continuous sediment movement, which prevents tree growth and associated sedimentation of ﬁne material. The case-study of the Ise`re River (France) introduced by K. Richards in the last GBR conference and by J.L. Peiry in this one, is relevant in this context. The river, which was embanked two centuries ago, still maintained well-developed unvegetated bars within its straight channel. Following a peak ﬂow decrease due to hydroelectric production and bedload trapping upstream for the reduction of ﬂood risk, the mobile gravel bars of the Ise`re River have been tremendously encroached by riparian vegetation. This has created two distinct units within the embanked corridor: a ﬂoodplain forest and a permanent ﬂow channel.

Reference Habersack, H., Formann, E., Muhar, S., et al., 2007. Hydraulic, river morphological and ecological effects of river bed widenings, Mitteilungen Nr. 200 VAW-ETH Zurich, pp. 119–130.

Gravel-Bed Rivers VI: From Process Understanding to River Restoration H. Habersack, H. Pie´gay, M. Rinaldi, Editors r 2008 Elsevier B.V. All rights reserved.

739

28 Uncertain restoration of gravel-bed rivers and the role of geomorphology David A. Sear, Joseph M. Wheaton and Stephen E. Darby

Abstract River restoration projects in gravel-bed rivers are becoming increasingly sophisticated and complex as river managers and scientists attempt to deliver the goals of catchment-scale ecosystem restoration. With increased sophistication, come the dual challenges of recognizing and responding to the uncertainty inherent in the restoration process. Uncertainty is rarely explicitly recognised in current restoration projects and, where it is, the scope and deﬁnition are limited. In this paper we argue that uncertainty is a fundamental element of river restoration and that the sources of uncertainty are varied. A typology for understanding and communicating uncertainty in terms of these sources is presented. One of the myths surrounding uncertainty is the notion that being uncertain is the same as not knowing anything. In fact, when uncertainty is expressed as a statement of plausible outcome and/or signiﬁcance, expressing uncertainty is a very informative statement of knowledge. The signiﬁcance of uncertainty is explored conceptually and quantiﬁed for two contrasting examples from two gravel-bed river restoration projects. Respectively, these demonstrate that uncertainty in the conceptual model applied to a restoration project can have signiﬁcant impacts on the restoration process and that unreliability uncertainties can affect the design of bankfull channel dimensions. The paper concludes with a discussion of the approaches to incorporating uncertainty in river restoration projects, and argues for one that embraces uncertainty. We present an approach for embracing geomorphic uncertainty in physical habitat restoration, that uses coupled habitat and landscape evolution models to deﬁne the plausible outcomes for a given restoration project. 1.

Introduction

River management philosophy over the past decade has converged on the over-arching ethos of sustainable management of water and associated ecosystem functions within the spatial context of the river catchment (Graf, 2001; Ward et al., 2002; Downs and E-mail address: [email protected] (D.A. Sear) ISSN: 0928-2025

DOI: 10.1016/S0928-2025(07)11162-7

740

D.A. Sear, J.M. Wheaton, S.E. Darby

Gregory, 2004). Embedded within these ethos are the traditional river management aims of protection against damaging ﬂoods and the provision of adequate water supply to the human populations within the catchment (Fleming, 2002). However a range of ‘‘new’’ concerns have also emerged that relate explicitly to the functioning of aquatic ecosystems. These include preservation of physical integrity (Graf, 2001; Everard, 2004), restoration of ecosystem functions (Richards et al., 2002) and management of water, nutrient and sediment ﬂuxes at the catchment scale. As a result, approaches to river management are now more holistic in their conceptualisation (Everard and Powell, 2002; Newson, 2002), multidisciplinary (Fleming, 2002) and participatory in their implementation (Clark, 2002). River restoration is a key component of this new river management, widely seen as the process through which river basin management’s aspirations and targets will be delivered (NRC, 1999; European Council, 2000). Thus, a tremendous diversity of river restoration projects have been undertaken throughout the world (Boon et al., 1992; Calow and Petts, 1994; Brookes and Shields, 1996; Iversen et al., 1998; Malakoff, 2004) in response to a well-documented array of adverse impacts from anthropogenic disturbances (Brookes, 1988; Coltorti, 1997; Graf, 2001; Knox, 2001). The restoration science community has responded to the large demands (from the practitioner, policy-maker and stakeholder communities) for ways to restore and mitigate such problems with a rich assortment of approaches, strategies and tools (e.g., Sear, 1994; Brookes, 1995; Brookes and Sear, 1996; FISRWG, 1998; Wissmar and Beschta, 1998; Gilvear, 1999; Koehn et al., 2001). As restoration evolves, projects are becoming more expensive, complex and technically difﬁcult, with lifetimes now extending over geomorphologically relevant timescales (Newson, 2002; Sear and Arnell, 2006). With increasing sophistication comes additional risks in terms of setting and meeting realistic project targets and to communicate complex models of river environments to stakeholders. Indeed, the results from recent monitoring programmes are beginning to cast doubt on the ability of restoration projects, as currently practised, to deliver some of these targets (Harrison et al., 2004; Williams et al., 2004). Central to progressing more sophisticated models of river restoration is our ability to comprehend and communicate the uncertainty in the science to a stakeholder base (that may include other scientiﬁc disciplines) that has been brought up with the notion that environments can be managed in a deterministic fashion. Uncertainty exists throughout the restoration process (Fig. 28.1a), yet paradoxically, most restoration projects fail to explicitly identify or communicate the uncertainty (Wheaton et al., 2006). The scientiﬁc community implicitly accepts, and to a certain extent thrives on, the inherent uncertainties in conceptual ideas, approaches, tools and strategies it provides (Pollack, 2003). However, restoration science has largely failed to transparently communicate these uncertainties to restoration practitioners, policy-makers and stakeholders (Wissmar and Bisson, 2003). Presumably, either adaptive management (encouraging the treatment of restoration projects as ‘learning by doing’) or a lack of long-term monitoring have prevented the emergence of any major consequences from ignoring uncertainties (Walters, 1997; Clark, 2002). Should long-term monitoring actually take place and reveal a systematic pattern of project failure (e.g., Kondolf, 1995; Kondolf et al., 2001; Downs and

Uncertain restoration of gravel-bed rivers and the role of geomorphology

741

TIME (ENVIRONMENT SENSITIVE) -2 YEARS

RESTORATION PROCESS

UNCERTAINTY DRIVERS (partial list)

0

0 - 5 YEARS

5 - 50 YEARS

DESIGN

CONSTRUCT

POSTCONSTRUCTION ADJUSTMENT

AVAILABILITY OF ANALOGUES

PERSONNEL TRAINING

EQUILIBRIUM?

FUNCTIONAL AIMS (e.g. Flood Defence)

SITE CONDITOINS

FLOOD TIMING

CLIMATE CHANGE

ENERGY OF SYSTEM

CLARITY OF DESIGN

VEGETATION COLONISATION

OTHER RESTORATION EFFECT

ADJUSTMENT

CATCHMENT CHANGE

Practically Immeasurable

Inherent Randomness of Nature Value Diversity (Sociopolitical) Behavioral Diversity Societal Randomness Technological Surprise

B.

Conflicting Evidence UNCERTAINTY DUE TO VARIABILITY

Reducible Ignorance Indeterminacy Irreducible Ignorance

DEGREE OF UNCERTAINTY

Lack of Observations & Measurements

UNCERTAINTY DUE TO LIMITED KNOWLEDGE

STRUCTURAL UNCERTAINTY

Inexactness

UNRELIABILITY

A.

Increasing Uncertainty

Figure 28.1. Uncertainty conceptualised. (A) Sources of uncertainty in the restoration process. (B) Typology for uncertainty based on sources of uncertainty. (Adapted from Rotmans and Van Asselt, 2001.)

Kondolf, 2002), stakeholders and policy-makers may well demand explanations from restoration scientists and practitioners (Wilcock, 1997). If such a process also reveals the plethora of uncertainties in restoration objectives, strategies and approaches that have to date been ignored, then the community might well be accused of gross negligence. Or, more to the point, the restoration community may suffer funding cuts or be sued. To address these issues, we herein wish to provide a context for understanding uncertainty in gravel-bed river restoration. Our objectives in this paper are twofold. First, we shall demonstrate that a variety of sources of uncertainty exist in river

D.A. Sear, J.M. Wheaton, S.E. Darby

742

restoration and that their signiﬁcance is context speciﬁc (see Johnson and Rinaldi, 1998). Second, we shall contrast a variety of approaches to deal with uncertainty. From the outset, it is important to distinguish that our primary concerns about uncertainties as they relate to river restoration are not about the uncertainties themselves; but rather the implications of the failure of the restoration community to communicate them. The basic premise is that this leads to unrealistic expectations among stakeholders about the outcomes of restoration projects.

2.

The concept of uncertainty

Although uncertainty is a fundamental and basic concept in the sciences (Jamieson, 1996; Pollack, 2003), we are often careless with its semantics. This carelessness makes it easy to confuse uncertainty with a complete lack of knowledge. For our purposes, we will consider a broad deﬁnition of uncertainty proposed by Van Asselt (2000), in which uncertainty is deﬁned in terms of its source. Van Asselt (2000) considers uncertainty to stem from two sources – limited knowledge (as opposed to a lack of knowledge) and variability (Fig. 28.1b). Uncertainty due to variability includes natural variation, natural variability (spatial and temporal), non-linear, random and chaotic behaviour, as well as value diversity within a society (Rotmans and Van Asselt, 2001). Paradoxically, uncertainties due to variability contribute to uncertainty due to limited knowledge, which can be segregated into structural and unreliability uncertainties. Unreliability uncertainties are those most typically quantiﬁed by scientists (e.g., error in a measurement). However, structural uncertainties are typically of a higher degree than unreliability uncertainties and represent more fundamental barriers to understanding (Van Asselt and Rotmans, 2002). Conflicting evidence is one type of structural uncertainty, which is particularly prevalent in river restoration for deciding which approaches or what types of analogues to use. There exist also uncertainties due to reducible and irreducible ignorance, which represent the difference between things ‘we could but don’t know’, and things ‘we cannot know’ (Van Asselt and Rotmans, 2002). We will use this typology of uncertainty throughout the rest of this paper to identify the speciﬁc uncertainties discussed, and a description of the terms is provided in Table 28.1 to assist the reader. Where a speciﬁc source of uncertainty is classiﬁed according to the typology, we will use italics. Additionally, where we are generically interested in future outcomes (i.e., prediction) we will call the set of all plausible outcomes the ‘plausible outcome space.’

3.

Signiﬁcance of uncertainty

As pointed out in the introduction, uncertainties are ubiquitous in river restoration. If one assumes that uncertainty is unequivocally a negative attribute, this paints a very bleak outlook for river restoration. However, under the broader deﬁnitions of uncertainty described in the previous section (e.g., Van Asselt and Rotmans, 2002), to identify our uncertainty about something is actually a statement of our knowledge.

Uncertain restoration of gravel-bed rivers and the role of geomorphology Table 28.1.

743

Glossary of terms used in ﬁgure.

Uncertainty typology

Term

Description

Structural uncertainty

Irreducible ignorance Indeterminacy Reducible ignorance

Things we cannot know Things we will never know We do not know what we do not know – this is information/knowledge that is accessible but as yet we have not as a community discovered it. We do not know what we know – knowledge is inexact and different interpretations of the same phenomena/information exist. We know what we do not know – refers to phenomena or measurements that cannot as yet be undertaken perhaps because of scaling issues or the technology available at the time. Could have, should have, would haveybut did not. A real aspect in most projects is the pieces of information that we were unable to record or simply did not record. Related to error, imprecision and accuracy of the information/ measurements acquired.

Conﬂicting evidence

Unreliability uncertainty

Practically immeasurable

Lack of observations and measurements

Inexactness

Source: Adapted from text of Van Asselt and Rotmans (2002).

A pedagogic exploration of the semantics of uncertainty without considering the signiﬁcance of uncertainty is of limited applied value. The signiﬁcance of our uncertainty depends entirely on the context and perspective from which we consider it. For example uncertainty in nutrient loading may be an eminent concern in a system being restored for water-quality purposes; yet may be of little consequence in the same system being restored for ﬂood protection purposes. Where restoration is driven by a narrow set of stakeholder, agency or societal objectives (e.g., single species salmonid restoration) the signiﬁcance of uncertainty is relatively simpliﬁed. Increasingly, however, restoration practitioners have been considering more sophisticated, multipurpose objectives simultaneously (catchment restoration, ecosystem restoration, dynamic system restoration, etc.) With more sophisticated restoration objectives come additional sources of uncertainty with interdependent signiﬁcance. Returning to the example above, if the system were being restored for both ﬂood protection and water-quality purposes, all the sources of uncertainty deemed to be signiﬁcant for either objective would have to be considered. If one were trying to communicate their uncertainty about nutrient loading to a group of water-quality scientists, reporting signiﬁcance in terms of nutrient concentrations would be perfectly acceptable. However, if one were communicating

744

D.A. Sear, J.M. Wheaton, S.E. Darby

the same uncertainty to a lay audience of stakeholders at the local ﬁshing club, a more effective means of communicating the signiﬁcance of that uncertainty might be in terms of the potential impacts of uncertain nutrient loading on ﬁsh. It is premature to say too much about whether uncertainty is signiﬁcant in restoration because there has been very little research on the topic (Wheaton, 2004; Wheaton et al., in press). Basic reasoning is adequate to identify the ubiquity of uncertainty in restoration. However, until the signiﬁcance of those uncertainties is considered, it is difﬁcult to say whether it matters or not. Unfortunately for those looking to make generalizations, signiﬁcance is a value-laden and context-speciﬁc consideration. Although this makes signiﬁcance a difﬁcult characteristic to generalise about, it also provides interpretations of uncertainty that can have greater practical utility. Hence, here we can only provide the reader with a couple of speciﬁc case studies of the signiﬁcance of uncertainty in river restoration. We can, however, draw a precautionary conclusion that based on the ramiﬁcations of the uncertainties we know to exist, the plausible outcome space of restoration activities certainty includes outcomes that could do more harm than good. Regrettably, the restoration community has instead chosen to assume that because its intentions are ‘good’, the outcomes cannot be worse than the status-quo. This should not be misinterpreted as a general scepticism about restoration. Rather, as we show in the discussion, it means that restoration science should embrace uncertainty in order to help the restoration community and general public form more realistic expectations about restoration. First, we will explore these concepts with two speciﬁc examples common to many restoration projects: (a) conceptual models and (b) reference conditions. 3.1.

Knowledge uncertainty and the role of the conceptual model in river restoration

Conceptual models of how river systems function lie at the heart of restoration projects. They are used to inform the understanding of river system function and river system form. In turn, these affect the design and management of the restoration project. Wheaton et al. (2004) argued that numerous conceptual models exist within the scientiﬁc literature that might be selected for a given restoration project; but, they advocated using these only to develop multiple working design hypotheses that could be tested and reﬁned prior to constructing a restoration project based on them. Rutherfurd et al. (in press) concluded that however powerful the conceptual model may be, it brings with it high risks if it has poor validity or is just plain wrong with respect to the given river system. This notion is not unfamiliar; for example Lewin et al. (1998) argued against the validity of applying regime models to gravel-bed rivers in an era of environmental change, preferring instead to take a more long-term view of river channel dynamics. In the van Asselt (2000) typology, Lewin et al.’s (1998) argument for extending the temporal scale of investigation of river management projects was intended to reduce uncertainty due to natural variability by quantifying the nature and drivers of natural variability. More recently, conceptual models that highlight the importance of natural variability have been embraced as an essential feature of river restoration projects (Wissmar and Bisson, 2003; Hughes et al., 2005). The example below illustrates some of the uncertainties and pitfalls of using conceptual models in river restoration.

Uncertain restoration of gravel-bed rivers and the role of geomorphology

745

The restoration of salmonid spawning habitat in the UK rivers is driven by a welldocumented decline in stocks (WWF, 2001). Despite widespread evidence for the importance of marine factors, river management agencies have applied a conceptual model based on the premise that recruitment is constrained by the quality of the incubation environment within the freshwater phase of the salmon life cycle (Reiser, 1998). This is not surprising when you consider that river management agencies have little or no jurisdiction in the marine environment. The conceptual model correlates ﬁne sedimentation of spawning gravels directly with egg survival (Fig. 28.2a). Management response has been driven by this conceptual model to undertake investment in gravel cleaning, channel narrowing, bank erosion control and catchment-scale treatment of soil erosion, (Greig et al., 2005). Furthermore, considerable investment is made in assessing spawning habitat quality based on indicators that measure the proportions of ﬁne inorganic sediment within the gravels (Reiser, 1998). In the UK, a general absence of monitoring has led to the widespread assumption that these approaches to river restoration are adequate (McMellin et al., 2002) and meet the implicit restoration targets of increasing survival rates to swim up. Thus, a lack of monitoring means it is unknown whether salmon stocks are improving, declining further, or not responding at all to restoration efforts. This is an example of uncertainty due to reducible ignorance (refer to Fig. 28.1b), whose signiﬁcance arguably undermines both the effectiveness of past restoration and appropriateness of future restoration. From a scientiﬁc perspective, the recruitment-constrained conceptual model has been challenged by new research on the factors affecting the incubation of salmonids within UK river gravels (Greig et al., 2005). In a study of four contrasting river environments, Greig et al. (2005) documented widespread variability in salmon egg survival. These were attributed to a suite of factors, including sedimentation by inorganic material, oxygen depletion due to oxidation of organic sediments, reduced intragravel ﬂow rates due to low ﬂows, and potentially, decreased oxygen concentrations due to upwelling of groundwater with low dissolved oxygen content. The research also tested the effectiveness of a range of popular sediment-based indicators of incubation habitat quality utilised by ﬁsheries managers. The popular indicators were demonstrated to be ineffective since they failed to represent the complex factors leading to poor survival (Greig et al., 2005). The development of a new conceptual model from Greig et al.’s (2005) research highlights the multiple and interacting factors that can inﬂuence the survival of incubating salmonids (Fig. 28.2b). The improved scientiﬁc understanding arising from this model results not in a reduction in uncertainty due to limited knowledge, but rather a transformation from uncertainty due to reducible ignorance to uncertainties of lesser signiﬁcance (see Fig. 28.1b) and of more information value (Van Asselt and Rotmans, 2002):

uncertainties due to unreliability – current restoration projects do not measure the necessary variables (but we now know what they are) and the variables when measured are prone to measurement errors structural uncertainties due to conflicting evidence – empirical observations within the scientiﬁc literature strongly support the importance of inorganic sedimentation on salmonid egg survival (Reiser, 1998)

D.A. Sear, J.M. Wheaton, S.E. Darby

746

INFLUENCES ON OXYGEN SUPPLY: Sediment Deposition

Seepage Velocity [Sediment reduces pore space and gravel permeability]

MANAGEMENT RESPONSES:

Gravel Cleaning: pressure washing fine sediment from spawning gravels

Blocks Emergence Build up of ammonia

[Sediments create impermeable seal]

Channel Morphological Manipulation: to increase flow velocity to flush fines from within the gravel bed

Catchment Treatments: to reduce ingress of fine sediments delivered from soil erosion

Survival to emergence of salmon progeny

Survival to Hatching

Bank Erosion Control: to reduce fine sediment ingress.

A. Stream discharge Hydraulic gradient Stream geometry Scour Gravel shape, size and composition Sediment deposition (size and distribution) Oxygen concentration instream Stream water temperature Stage of embryo development Size of embryo

Spatial distribution of eggs within spawning gravels

Seepage velocity

Substrate permeability BOD of organic sediments

Presence of groundwater upwelling

Intragravel oxygen concentration

Oxygen supply to egg pocket

Emergence from gravel

Intragravel water temperature Rate of uptake of oxygen by embryos

Probability of embryo survival

B. Figure 28.2. Contrasting conceptual models of spawning incubation. (A) The conceptual model underpinning a suite of current restoration practices. (B) A conceptual model of spawning incubation that results in different potential causes than the currently adopted model.

uncertainty due to variability – inherent randomness of nature exhibited in the dynamics of physical and chemical processes

This so-called ‘transformation’ of uncertainty can be viewed as an increase in uncertainty, as further study only revealed further uncertainties. However, in terms

Uncertain restoration of gravel-bed rivers and the role of geomorphology

747

of the signiﬁcance of these uncertainties, the additional uncertainties actually focus attention on the most important factors of the incubation problem and act to provide helpful information. The original conceptual model (Fig. 28.2a) was introducing a structural uncertainty due to reducible ignorance as it was being applied. In other words, it was not known whether the model was correct, but in principle it could be known by testing the model. The testing of the conceptual model and reﬁnement of a new conceptual model is not free of uncertainty, but conﬁnes this limited knowledge to a much narrower plausible outcome space.

3.2.

Unreliability uncertainty and the design of reference condition channels

River restoration often requires some degree of channel design. The signiﬁcance of uncertainty in the design process is seldom quantiﬁed, and attitudes to uncertainty vary depending on the level of accuracy demanded by the project. Many restoration projects (e.g., those that involve signiﬁcant changes in ﬂood risk as is the case when a channel is reconnected to its ﬂoodplain) might require considerable design accuracy to ensure an outcome that satisﬁes stakeholders. We contend that there is a systematic distinction in the required design precision between upland and lowland gravel-bed river environments. In high-energy upland systems, rates of morphological adjustment are frequently sufﬁcient that designs based on ‘passive’ approaches represent cost-effective means of restoring or rehabilitating the functional characteristics of substantial lengths of river channel (Brookes and Sear, 1996). In comparison, lowland gravel-bed rivers systems often have a combination of low stream power and resistant channel boundaries that do not allow natural recovery in anything less than unacceptably long timescales (Brookes and Sear, 1996). In such cases the legacy of channel design mistakes would, without corrective intervention, last for many years, implying that a high accuracy in the restoration design is required, as well as in the implementation of that design. Moreover, lowland systems are frequently heavily disturbed and are often those systems most in need of restoration (Sear et al., 2000). A restoration scenario typical of lowland river restoration in the UK was investigated to quantify the magnitude of uncertainty in the ﬁnal design arising from unreliability uncertainties due to inexactness and lack of observations and measurement. The river Cherwell is a low gradient, gravel-bed river conﬁned within cohesive ﬂoodplain alluvium. The Cherwell has a long history of modiﬁcation to its hydrological regime, surrounding catchment land cover and management, planform, cross-section and long proﬁle (for the complete site description and study, see Sear et al., 2001). The restoration design for the Cherwell was theoretical, and formed part of a hydrological modelling study to determine the effectiveness of different channel and ﬂoodplain restoration scenarios on downstream ﬂood peaks. An element of this wider project necessitated the reconstruction of the channel dimensions and ﬂoodplain vegetation structure prior to major channel modiﬁcations and land management in the 19th century. The initial approach was to identify semi-natural reference sites from which an empirical hydraulic geometry model might be developed. This method itself is prone to structural uncertainties that question its validity. Assuming that reliable data are available (unreliability uncertainties), power functions relating channel geometry to drainage

748

D.A. Sear, J.M. Wheaton, S.E. Darby

area can therefore be developed and used to predict values for any given point in the river network. In the case of the river Cherwell, values of discharge and depth for the pre-disturbed watercourse were not available due to the absence of suitable analogue reaches and historical data sources. Paradoxically, this condition is typical of heavily disturbed lowland gravel-bed rivers. As a way around this, we estimated the predisturbance channel dimensions with a combination of empirical methods and analytical modelling techniques. Given the methodological uncertainties in the work-around method, we explored their signiﬁcance in terms of the designed channel dimensions. 3.2.1.

Estimating pre-disturbance bankfull channel width

Values of channel width were derived with a relatively small degree of inaccuracy uncertainties from large-scale historic maps (Downward, 1995), with catchment area used as a surrogate for discharge (Sear, 1996). Downward (1995) deﬁnes two main categories of error in deriving map-based estimates of channel boundaries: inherent errors and operational errors. Inherent errors are included at each point in the process of data capture from landscape to ﬁnal visualisation on the map (unreliability uncertainties), including those arising from the surveying process itself. Downward (1995) provides a method for estimating the component of inherit error arising from the process of registering the paper map into digital format (unreliability uncertainties); but, since we were not comparing locations of channel boundaries at different points in time, positional accuracy were not relevant. It was assumed that inherent error, the result of survey and cartographic errors at the time of map production (and the source of structural uncertainty in the method) was largely unquantiﬁable. Unreliability uncertainties in the form of operational errors occur in the estimation of channel width from the raster map. These arise from the pixel representation of the channel boundary. For this study, the pixel area was set to 1 m2 since this corresponded to the approximate thickness of the lines denoting channel boundaries on the original map. Representation of the channel boundary usually involved only one pixel, but where channel curvature was signiﬁcant, this increased to two pixels. The centre of each pixel, or the boundary between two pixels, was taken as the location of the channel boundary. Using this standard approach, the channel width could therefore be said to lie within 71.0 m for locations where the channel boundary was represented by a single pixel, or 72.0 m for locations where the line was represented by two pixels. Quantiﬁable total mapping error (TME) was therefore estimated to be 71.0 to 2.0 m. This value represents the summation of all recognised unreliability uncertainties in the method. To provide estimates of bankfull width for the preengineered channel, nine sites were identiﬁed that had a natural planform. At these sites, a length of channel of up to 200 m was deﬁned, and 15 channel widths were digitally measured using the on-screen distance tool. The channel widths were divided into bends and inﬂexion points. The error bands shown on Fig. 28.3a are, however, substantial with average errors of 71.3–2.2 m. 3.2.2.

Estimating pre-disturbance bankfull channel depth

Having deﬁned pre-engineered width values, channel depth was estimated using a rational regime type model, VARSLOPE (Millar and Quick, 1993, 1998).

Uncertain restoration of gravel-bed rivers and the role of geomorphology

749

Width (m)

A 10

1 10

100

1000

Depth (m)

B 1

10

100

1000 2

Drainage Area (km ) Figure 28.3. Uncertainty bounds around estimations of channel bankfull depth and width for the predisturbance river Cherwell.

VARSLOPE simulations were undertaken at speciﬁc sites in the restoration reaches deﬁned previously. Accordingly, it was necessary to deﬁne the values of model input parameters at these locations. Most parameter values were deﬁned in relation to direct physical measurements (e.g., bed-material grain size, bank material characteristics), but others require certain assumptions to be made, thereby introducing additional uncertainty into the analysis both via unreliability uncertainties based on the values used in the estimations, and structural uncertainties that are inherent (and often unspeciﬁed) in the methods adopted. For example the bankfull discharge is commonly assumed to be a formative discharge with a recurrence interval (RI) in the range of 1–2 years (Wolman and Leopold, 1957). In this study, we used a 1.5-year RI to deﬁne the formative ﬂow, though we have no means of verifying that this value is appropriate. Furthermore, no data were available to deﬁne the Cherwell’s sediment load in the study reaches. We estimated this parameter by constructing an empirical relationship between ﬂow discharge and sediment load using data from Hey and Thorne (1986). The Hey and Thorne (1986) data are based on the physical characteristics of 62 British gravel-bed rivers, but we restricted our analysis to a subset of the data for rivers with similar gradients (o 0.0015) to the Cherwell. Certain assumptions are also implicit in our use of well-established empirical–physical relationships (Hey, 1979; Clifford et al., 1992) to calculate bed roughness height based on measured bedmaterial grain sizes (Table 28.1). Finally, we assumed that the bank roughness height was equal to that calculated for the bed. In addition to uncertainty introduced by these assumptions, one input parameter (the critical shear stress for entrainment of bank material, tc) is not readily measurable, nor can it easily be estimated, except via

750

D.A. Sear, J.M. Wheaton, S.E. Darby

model calibration. This was achieved by adjusting tc until simulated equilibrium width values agreed with the ‘observed’ bankfull width derived from the empirical hydraulic geometry analysis undertaken previously (Fig. 28.3). Since the calibration process forces agreement with ‘observed’ data, uncertainty associated with each of the individual input parameters described previously is effectively lumped into a single term (tc). This is convenient because, if the gross uncertainty in tc can be established, the impact of this on simulated depth can readily be determined. Speciﬁcally, multiple simulations can be undertaken using a suitable range of tc values obtained from calibrations using width values according to the quantiﬁed uncertainty. The results can then be used to deﬁne a corresponding range of simulated depth values. In a preliminary set of VARSLOPE simulations we determined the uncertainty in tc by calibrating the model using a range of channel width values reﬂecting the (map-derived) error quantiﬁed for each site. Subsequently, a second set of simulations was undertaken to evaluate the impacts of uncertainty in tc on simulated pre-disturbance depth values (for complete reporting, see Sear et al., 2001). Based on the preceding analysis, pre-disturbance channel dimensions of the river Cherwell can be reconstructed as a function of drainage area and visualised in relation to the associated degree of unreliability uncertainty. The relative uncertainty is greater for the reconstructed width (739– 724%) than the depth (728–720%), but for both variables the uncertainty is scale-dependent, declining in the stated ranges as drainage area increases from 150 to 800 km2 (Sear et al., 2001). This implies that unreliability uncertainty (e.g., inaccuracy) is not in this case ampliﬁed as a result of using uncertain channel width data in the calibration of the VARSLOPE model. The decline in relative uncertainty as a function of increasing width and depth as drainage area increases is explained by recalling that uncertainty in this case study is essentially derived from the ﬁxed (73.25 m) error involved in estimating channel width from the map data. These, admittedly, speciﬁc types of uncertainties directly inﬂuence speciﬁcation of the designed channel dimensions. One can visualise the uncertainty reported in Fig. 28.3 as the dimensions of a cross-section with a vaguely deﬁned bed and bank boundary. The signiﬁcance of this vaguely sized channel can be expressed simply as a comparison between ﬂoodplain reconnection and downstream ﬂooding at the two extremes of channel capacity resulting from the width and depth uncertainty (e.g., smallest width and depth combination versus largest width and depth combination). Were the aim of the Cherwell restoration project habitat enhancement instead of ﬂoodplain reconnection and ﬂood defence, the signiﬁcance would be expressed in terms of inﬂuence on physical habitat. Depending on the metric of signiﬁcance deemed important, the same uncertainty (in this case channel dimensions) could be essential to project success, or have virtually no signiﬁcance.

4.

Contrasting philosophies towards uncertainty

Having established that uncertainties are prevalent in restoration and that their signiﬁcance varies depending on the context, we are still left with a philosophical

Uncertain restoration of gravel-bed rivers and the role of geomorphology

751

choice as to how to approach uncertainty. There are at least ﬁve philosophical approaches towards uncertainty (Wheaton et al., in press):

Ignore Uncertainty: A passive approach to uncertainty that works if uncertainty turns out to be insigniﬁcant. Uncertainty might be ignored if it has been proven to be insigniﬁcant. More often, uncertainty is ignored because we are ignorant to it or aware of it but do not have the time, resources or know-how to deal with it actively. Eliminate Uncertainty: Only considers a narrow deﬁnition of uncertainty and assumes that an absolute answer exists to the thing(s) we wish to know (e.g., someone’s name is an example of a something that uncertainty can be completely eliminated about). Under broader deﬁnitions of uncertainty (e.g., Van Asselt and Rotmans, 2002), this approach is typically impossible. The philosophy is based on a positivist view of uncertainty as an absolute lack of knowledge and/or sign of weakness. Reduce Uncertainty: This is a prevalent approach amongst reductionist scientists, in which uncertainty is viewed as an unfortunate characteristic of the things we wish to know about. Under the Van Asselt and Rotmans (2002) typology, this approach only makes sense for unreliability uncertainties (e.g., measurement errors). The approach accepts that some uncertainty will always be present, but strives to reduce it to an absolute minimum. Cope with Uncertainty: This is a slightly more practical adaptation of the reduce uncertainty philosophy, in which there is a fuller admission of the prevalence of uncertainty. The approach, therefore, only seeks to reduce uncertainty where it is practical to, and learn to adopt ways of coping with uncertainty otherwise. The approach still fundamentally views uncertainty as a negative thing. Embrace Uncertainty: This approach only makes sense if a very broad deﬁnition of uncertainty is accepted (e.g., Van Asselt and Rotmans, 2002). Uncertainty is viewed without contempt or admiration, and it is recognised that it sometimes results in positive outcomes (e.g., surprise, unforeseen beneﬁts) and sometimes results in negative outcomes. Uncertainty is embraced in that it is accepted for what it is – information. In only extreme cases is uncertainty a complete lack of knowledge (e.g., irreducible ignorance), and in most cases uncertainty is thought to be transformable into uncertainties of lesser magnitude that provide more information.

The overlap between these ﬁve philosophical strategies is illustrated in Fig. 28.4. To ‘ignore’ uncertainty is the most basic of the philosophical strategies towards uncertainty, and the one most commonly exercised by the restoration community (Wheaton et al., 2006). The appropriate choice of approach depends very much on the context of the restoration approach, the signiﬁcance of the uncertainty under consideration and the adopted deﬁnition of uncertainty. Regardless of the approach, the signiﬁcance of essential uncertainties should be assessed and more clearly communicated to the restoration community. Efforts based strictly on reducing uncertainty demand enhanced levels of monitoring and site investigation to establish causation, but these are likely to be frustrated by uncertainty due to variability. In the Cherwell example, the level of uncertainty was quantiﬁed and the sources constrained. Approaches based

D.A. Sear, J.M. Wheaton, S.E. Darby

752 Historical Views of Uncertainty

Contemporary Views of Uncertainty

Progressive View of Uncertainty

REDUCE UNCERTAINTY IGNORE UNCERTAINTY

COPE WITH UNCERTAINTY ELIMINATE UNCERTAINTY

EMBRACE UNCERTAINTY

“UNCERTAINTY IS”: “Not Acknowledged”

“Negative”

“a Reality”

“Potentially Positive”

Figure 28.4. Five philosophical attitudes towards uncertainty. The Venn diagram is meant to illustrate the overlap between contemporary attitudes towards uncertainty. Note that ignoring uncertainty, shares no overlap with contemporary attitudes towards uncertainty. (Figure taken from Wheaton, 2004.)

on further constraining uncertainty will again require more investment in pre-project research and investigation. Ignoring the uncertainty may invalidate the model output, which in the Cherwell example could have implications for the understanding and management of downstream ﬂooding. If the ﬂooding is acceptable, a more relaxed approach to dealing with the uncertainty may be warranted. Incorporating uncertainty speciﬁcally within a modelling context is an established method for coping with uncertainty, and several approaches exist to support this; for example GLUE (Beven, 1996; Brazier et al., 2000) or Monte Carlo simulation (Binley et al., 1991). In almost any restoration example, the uncertainty in the project outcome (e.g., uncertainty due to natural variability arising from subsequent channel adjustment) will be critical in determining whether the project is cast as a failure or a success. As stressed in the introduction to this paper, an unforeseen project outcome may be identiﬁed by stakeholders as project failure, rather than success (Kondolf, 1998). Here it is helpful to revisit this concept of plausible outcome space. In Fig. 28.5 all outcomes are deﬁned within the space bounded by the four shaded boxes. The boxes represent the four variants of outcomes based on desired versus undesired outcomes and unforeseen beneﬁts versus unforeseen consequences. Notice that the plausible outcome space only occupies a smaller subset (represented here with a circle) of the larger outcome space. If we do nothing other than estimate the range of plausible outcome space, our uncertainty is contained within this space. If we fail to communicate our uncertainty, we risk encouraging the incorrect presumptions that the target state will certainly be achieved or that our uncertainty occupies the entire output space. Within the restoration context, perhaps the most promising and practical philosophy towards uncertainty is to embrace it. Wheaton (2004) details the many reasons for this; but one of the most compelling is that under a broad and holistic consideration of uncertainty, it can be quite a positive thing (Clark, 2002; Lempert et al., 2003). From an engineering perspective it may be disappointing that our simulation models cannot make predictions with absolute certainty. However, uncertainty in

Uncertain restoration of gravel-bed rivers and the role of geomorphology

753

Undesired Outcome

OUTCOME SPACE (BOX)

TYPE TYPE B B

TYPE TYPE D D

TYPE A

TYPE C

Desired Outcome

UNCERTAINTY

Unforseen Benefits Unforseen Consequences PLAUSIBLE OUTCOME SPACE (CIRCLE) Figure 28.5. A mapping of the plausible outcome space.

deﬁning future scenarios to drive such models can explain why this is the case (Lempert et al., 2003). From a practical stand point, systems that exhibit natural variability in their processes are likely to be more self-sustaining and resilient (Bilbly et al., 2003; Wissmar et al., 2003). Furthermore, an open and transparent communication of uncertainties in restoration is likely to lead to more realistic expectations about the potential outcomes of restoration projects (Wheaton, 2004). The lack of an adaptive management approach to river management is likely to impede such communication (Newson and Clark, in press).

5.

Where does embracing uncertainty take us?

An interesting question arises about whether embracing uncertainty necessarily leads to more complicated explanations. The two case studies we have provided seem to suggest to the reader that the answer is yes. However, when the signiﬁcance of uncertainty is cast in a slightly different context, we can ﬁnd examples where embracing uncertainty need not necessarily be equated to seeking more complicated models (whether they are conceptual or mathematical). If we are willing to accept a broader range of plausible outcomes (e.g., a larger diameter circle in Fig. 28.5) we can adopt a simpler approach to embracing uncertaintyywe live with it! This is different to ignoring it. The expectation based on

754

D.A. Sear, J.M. Wheaton, S.E. Darby

ignoring uncertainty is currently that we expect the restoration to attain its prescribed target (e.g., type A or C in Fig. 28.5) – if it does not (e.g., type B or D in Fig. 28.5) then the restoration is a failure and unforeseen additional management will be necessary. The expectation based on embracing uncertainty would be that we would hope that the prescribed target is attained, but we fully accept that it might not be. In this case because we are explicit about the signiﬁcance of uncertainty, not attaining the initial target is not failure of the project but attainment of one of the other plausible end states. Other plausible end states can include both desirable outcomes and undesirable outcomes. However, it is important to recognise that unforeseen beneﬁts can arise from what initially might have been perceived as undesirable outcomes (e.g., type B in Fig. 28.5). Under a model of restoration wherein uncertainty is ignored, such undesirable outcomes would have been written off as failures. The beneﬁts of such failures are only recognised through monitoring and adaptively managing on this information. Conversely, unforeseen consequences can arise from what initially might have been perceived as a desirable outcome (e.g., type C in Fig. 28.5). Under a model that ignores uncertainty, such consequences would have gone undetected. To return to the example of spawning habitat; Wheaton et al. (2004) documented an unforeseen consequence in a project that met its restoration objectives and was assumed to be a success. Rehabilitated spawning beds downstream of a reservoir were constructed according to a design speciﬁcation and shown through monitoring and habitat suitability modelling (considering hydraulics) to be providing suitable spawning habitat. However, a few years after construction an unforeseen consequence of reservoir and hatchery operations was resulting in the smothering of the spawning beds with invasive growth of aquatic vegetation. This growth was not incorporated in the original modelling that supported the design. In an embracing uncertainty framework, this scenario might have been identiﬁed speciﬁcally as one of the plausible outcomes. Even if it were not foreseen, the adaptive management response focuses attention on the newly acquired understanding of the role of vegetation, rather than locking the managers into a programme of vegetation maintenance in order to achieve a static target. Under an ignoring uncertainty approach, all the valuable information gleaned from the restoration project, the high-quality spawning habitat provided in the early stages of the project’s lifetime, and the potential for higher quality habitat to return would be brushed aside and the project labelled as a failure. The acceptance of a range of plausible outcomes to a restoration project rather than a single target end state is supported by notions of ﬂuvial systems as dynamic in the face of changing external and internal drivers (climate change, land cover change tectonic activity etc.). Indeed, geomorphology is replete with examples and conceptual models that demonstrate more or less sensitivity to changes in driver variables. Furthermore, the historical record of river system response is seldom static. Moreover, ecosystem models recognise the value and importance of disturbance as one of the main drivers of biodiversity and functionality. Thus by embracing uncertainty, restoration projects would incorporate dynamic systems whilst recognising that a range of plausible outcomes exist that may be more or less beneﬁcial to ecosystem functioning, and which may have a range of consequences. An important element to this discussion is the consideration of spatial scale. The plausible outcome space in a restoration project is strongly related to a location.

Uncertain restoration of gravel-bed rivers and the role of geomorphology

755

Most restoration projects are small-scale, reach-based schemes, nested within a wider, changing catchment. Frameworks that embrace uncertainty need to incorporate the spatial dimension; for example the attainment of a desirable outcome may occur, but not within the physical space that was the initial focus of the restoration (e.g., the local community may beneﬁt rather than the river per se). Similarly, it might be possible to attain the desirable outcome elsewhere in the system rather than at that speciﬁc location. As the requirements of restoration projects become more sophisticated and include, for arguments sake, a deﬁnition of habitat disturbance frequency and type; then the mapping of the plausible outcome space may need to be more reﬁned. The demands of the restoration science community then becomes one of providing methods for determining the plausible outcome space for a given set of boundary conditions and restoration design (i.e., ﬂuvial geomorphology). A further caveat is that these need to be spatial and capable of working over the longer timescales associated with ecosystem restoration. One approach being explored is to utilise the growing ﬁeld of landscape evolution modelling (LEM), and explicitly adapt it for the forecasting of plausible outcomes from single or multiple restoration projects within a catchment. Recent advances in landscape scale modelling are providing opportunities to integrate hydrological, land use history and geomorphology with models of vegetation succession (Richards et al., 2002). Coulthard and Macklin (2001) report such an exercise for a medium-sized basin in the UK using the (cellular automaton evolutionary slope and river (CAESAR) model. Such models offer the opportunity for river managers and stakeholders to engage with concepts of dynamic natural environments such as complex response, and enable informed discussion and communication of the signiﬁcance of geomorphic adjustment over longer timescales (Sear and Arnell, 2006). Furthermore, spatially distributed models of hydrology and sediment ﬂux share commonalities with landscape ecology models of patch dynamics. These provide a basis for more effective long-term simulation of ecosystem processes. We are currently reﬁning the CAESAR model to include an ecohydraulic model within it and using the simulations to explore the plausible outcome space of restoration activities across a case study catchment in California’s coast range. The process of modelling individual simulations that make predictions of a single plausible outcome is modularly implemented and diagrammed in Fig. 28.6. The modular ﬂexibility of the simulation approach allows process and scenario variants to be used to develop many individual simulations whose outputs can be collectively analysed to explore the plausible outcome space. While such an intensive modelling effort may not be necessary or feasible in all restoration instances, it is being pursued here for two primary reasons. First, in situations where we would like to narrow down the possibility of a type D outcome and better understand the triggers and controls that lead to a type D outcome, scenario modelling shows some promise. Second, even though we might know that the uncertainty in restoration outcome is large (i.e., the plausible outcome space is a bigger circle) we really do not yet understand the signiﬁcance or ramiﬁcations of the uncertainty. Where ﬁsh, bank erosion and ﬂooding are the concerns, an LEM coupled to a ﬁsh habitat model can allow us to explicitly explore the signiﬁcance of the uncertainty.

D.A. Sear, J.M. Wheaton, S.E. Darby

756

Figure 28.6. An example of a landscape evolution modelling approach being used under the embracing uncertainty philosophy to address geomorphic uncertainties in river restoration.

6.

Conclusions

This paper has argued that the increasingly complex goals and targets set by the river restoration community of scientists and practitioners necessitate greater consideration of uncertainty. To date limited recognition of the uncertainties involved and certainly limited uptake in restoration planning (with notable exceptions) has been the practice. However, with the results of science-based monitoring projects beginning to question the validity of the restoration process as currently practiced, the signiﬁcance of uncertainty is becoming more apparent. Uncertainty is often crudely deﬁned, thus a fundamental step is to structure the sources of uncertainty within river restoration. The typology of Van Asselt (2000) is advanced as a suitable framework for understanding and identifying the sources of uncertainty within river restoration. The signiﬁcance of uncertainty is explored for two areas of the restoration process. The uncertainty in the conceptual models used to support the development of restoration strategies are shown to result in the failure to treat the actual source of the biological problem. In another example, the choice of conceptual models available leads to signiﬁcant design uncertainty. For some sources of uncertainty, it is possible to estimate the magnitude. This is undertaken for a typical restoration design problem using standard geomorphological modelling approaches. The resulting uncertainty around the estimation of pre-disturbance bankfull channel dimensions range from 20 to 39%.

Uncertain restoration of gravel-bed rivers and the role of geomorphology

757

It is demonstrated that uncertainty in restoration exists in a variety of forms that inﬂuence the whole of the restoration process; but these can be identiﬁed. The question remains how to respond to them. We argue that it is no longer appropriate to ignore uncertainty, and, through recognition of the multiple types and sources uncertainty it is impossible to constrain. We make a case instead for a route that seeks to embrace uncertainty, and we demonstrated both complicated and simpliﬁed examples of such. We also brieﬂy highlighted an active area of research that is seeking to embrace uncertainty using coupled LEM and habitat modelling. All embracing uncertainty approaches are implicitly linked to adaptive management approaches. These recognise from the outset that river systems are inherently dynamic and that it is the operation of these dynamics that form the basis for any restoration. Although absolute prediction of restoration outcomes is impossible due to uncertainty, scenario modelling might be used to gain new insight into what restoration may mean. The hope is that ultimately this will help decision-makers, restoration practitioners and stakeholders form more realistic expectations about restoration, as well as furthering the scientiﬁc understanding of river response in a changing world.

Acknowledgements We would like to thank the EU LIFE ‘‘Wise Use of Floodplains’’ Project for funding the Cherwell portion of this research. Dr Rob Scaife provided the palaeoenvironmental reconstruction of the Cherwell ﬂoodplain, while Dr Sally German and the GeoData Institute undertook the cartographic (GIS) analysis and visualisation. The second author is supported by funding from the School of Geography at the University of Southampton, the Centre for Ecology and Hydrology and an Overseas Research Studentship. All this support is highly appreciated.

References Beven, K., 1996. Equiﬁnality and uncertainty in geomorphological modelling. In: Rhoads, B.L., Thorn, C.E. (Eds), The Scientiﬁc Nature of Geomorphology. Proceedings of the 27th Binghamton Symposium in Geomorphology. Wiley, Chichester, UK, pp. 289–314. Bilbly, R.E., Reeves, G.H., Dolloff, C.C., 2003. 6. Sources of variability in aquatic ecosystems: factors controlling biotic production and diversity. In: Wissmar, R.C., Bisson, P.A., and Duke, M. (Eds), Strategies for Restoring River Ecosystems: Sources of Variability and Uncertainty in Natural and Managed Systems. American Fisheries Society, Bethesda, MD, pp. 129–146. Binley, A.M., Beven, K.J., Calver, A., Watts, L.G., 1991. Changing responses in hydrology – assessing the uncertainty in physically based model predictions. Water Resour. Res. 27 (6), 1253–1261. Boon, P.J., Calow, P., Petts, G. (Eds), 1992. River Conservation and Management. Wiley, Chichester, UK, 470pp. Brazier, R.E., Beven, K.J., Freer, J., Rowan, J.S., 2000. Equiﬁnality and uncertainty in physically based soil erosion models: application of the glue methodology to WEPP – the water erosion prediction project – for sites in the UK and USA. Earth Surf. Process. Landf. 25 (8), 825–845. Brookes, A. (Ed.), 1988. Channelized Rivers: Perspectives for Environmental Management. Wiley, Chichester, UK, 336pp.

758

D.A. Sear, J.M. Wheaton, S.E. Darby

Brookes, A., 1995. River channel restoration: theory and practice. In: Gurnell, A. and Petts, G.E. (Eds), Changing River Channels. Wiley, Chichester, UK, pp. 369–388. Brookes, A., Sear, D., 1996. Geomorphological principles for restoring channels. In: Brookes, A. and Shields, F.D. (Eds), River Channel Restoration: Guiding Principles for Sustainable Projects. Wiley, Chichester, UK, pp. 75–101. Brookes, A., Shields, F.D., 1996. River channel restoration: guiding principles for sustainable projects. Wiley, Chichester, UK, 433pp. Calow, P., Petts, G.E. (Eds), 1994. The Rivers Handbook: Hydrological and Ecological Principles. Vol. 2, Blackwell Scientiﬁc Publications, Oxford, England, 523p. Clark, M.J., 2002. Dealing with uncertainty: adaptive approaches to sustainable river management. Aquat. Conserv. Mar. Freshwat. Ecosyst. 12, 347–363. Clifford, N.J., Robert, A., Richards, K.S., 1992. Estimation of ﬂow resistance in gravel-bedded rivers: a physical explanation of the multiplier of roughness length. Earth Surf. Process. Landf. 17, 111–126. Coltorti, M., 1997. Human impact in the Holocene ﬂuvial and coastal evolution of the Marche region, Central Italy. Catena 30, 311–335. Coulthard, T., Macklin, M.G., 2001. How sensitive are river systems to climate and land-use changes? A model-based evaluation. J. Quat. Sci. 16 (4), 347–351. Downs, P., Gregory, K.J., 2004. River channel management: towards sustainable catchment hydrosystems. Arnold, London, 395pp. Downs, P.W., Kondolf, G.M., 2002. Post-project appraisals in adaptive management of river channel restoration. Environ. Manage. 29 (4), 477–496. Downward, S.R., 1995. Information from topographic survey. In: Grunnell, A.M. and Petts, G. (Eds), Changing River Channels. Wiley, Chichester, UK, pp. 303–323. European Council, 2000. Directive 2000/60/EC of the European Parliament and the Council of 23rd October 2000: establishing a framework for community action in the ﬁeld of water policy. Ofﬁcial J. Eur. Communities L327, 1–72. Everard, M., 2004. Investing in sustainable catchments. Sci. Total Environ. 324, 1–24. Everard, M., Powell, A., 2002. Rivers as living systems. Aquat. Conserv. Mar. Freshwat. Ecosyst. 12 (4), 329–337. FISRWG, 1998. Stream corridor restoration: principles, processes, and practices. GPO item no. 0120-A, Federal Interagency Stream Restoration Working Group, Washington, DC. Fleming, G., 2002. Learning to live with rivers – The ICE’s report to government. Proc. Inst. Civ. Eng. Civ. Eng. 150 (Special Issue 1), 15–21. Gilvear, D.J., 1999. Fluvial geomorphology and river engineering: future roles utilizing a ﬂuvial hydrosystems framework. Geomorphology 31 (1–4), 229–245. Graf, W.L., 2001. Damage control: dams and the physical integrity of America’s rivers. Ann. Assoc. Am. Geogr. 91, 1–27. Greig, S.M., Sear, D.A., Carling, P.A., 2005. Fine sediment accumulation in salmon spawning gravels and the survival of incubating salmon progeny: implications for spawning habitat management. Sci. Total Environ. 344, 241–258. Harrison, S.S.C., Pretty, J.L., Shepherd, D., Hildrew, A.G., Smith, C., Hey, R.D., 2004. The effect of instream rehabilitation structures on macroinvertebrates in lowland rivers. J. Appl. Ecol. 41, 1140–1154. Hey, R.D., 1979. Flow resistance in gravel-bed rivers. J. Hydraul. Div. ASCE 105 (4), 365–379. Hey, R.D., Thorne, C.R., 1986. Stable channels with mobile gravel beds. J. Hydraul. Eng. ASCE 112 (8), 671–689. Hughes, F.M.R., Colston, A., Mountford, J.O., 2005. Restoring riparian ecosystems: the challenge of accommodating variability and designing restoration trajectories. Ecol. Soc., 10(1), 12 [online] URL: http://www.ecologyandsociety.org/vol10/iss1/art12/ Iversen, T.M., Hansen, H.O., Madsen, B.L., 1998. River restoration – The physical dimension – Introduction. Aquat. Conserv. Mar. Freshwat. Ecosyst. 8 (1), 3–4. Jamieson, D., 1996. Scientiﬁc uncertainty: how do we know when to communicate research ﬁndings to the public? Sci. Total Environ. 184, 103–107. Johnson, P.A., Rinaldi, M., 1998. Uncertainty in stream channel restoration. In: Ayyub, B. (Ed.), Uncertainty Modelling and Analysis in Civil Engineering. CRC Press, Boca Raton, FL, pp. 425–437.

Uncertain restoration of gravel-bed rivers and the role of geomorphology

759

Knox, J.C., 2001. Agricultural inﬂuence on landscape sensitivity in the Upper Mississippi River Valley. Catena 42, 193–224. Koehn, J.D., Brierley, G.J., Cant, B.L., Lucas, A.M., 2001. River Restoration Framework, The National Rivers Consortium: Land and Water Australia, Canberra, ACT. Kondolf, G.M., 1995. Five Elements for effective evaluation of stream restoration. Restor. Ecol. 3 (2), 133–136. Kondolf, G.M., 1998. Lessons learned from river restoration projects in California. Aquat. Conserv. Mar. Freshwat. Ecosyst. 8 (1), 39–52. Kondolf, G.M., Smeltzer, M.W., Railsback, S.F., 2001. Design and performance of a channel reconstruction project in a coastal California gravel-bed stream. Environ. Manage. 28 (6), 761–776. Lempert, R.J., Popper, S.W., Bankes, S.C., 2003. Shaping the next one hundred years: new methods for quantitative, long-term policy analysis. The Rand Pardee Center, Santa Monica, CA. Lewin, J., Macklin, M.G., Newson, M., 1998. Regime theory and environmental change – irreconcilable concepts? In: White, W.R. (Ed.), International Conference on River Regime. Wiley, Chichester, UK, pp. 431–445. Malakoff, D., 2004. The river doctor: proﬁle Dave Rosgen. Science 305 (13 August), 937–939. McMellin, G., Walling, D.E., and Nicholls, D., 2002. Land use and ﬁsheries. Report W2-046/TR1, Environment Agency, Bristol, UK. Millar, R.G., Quick, M.C., 1993. Effect of bank stability on geometry of gravel rivers. J. Hydraul. Eng. 119, 1343–1363. Millar, R.G., Quick, M.C., 1998. Stable width and depth of gravel-bed rivers with cohesive banks. J. Hydraul. Eng. 124, 1005–1013. Newson, M.D., 2002. Geomorphological concepts and tools for sustainable river ecosystem management. Aquat. Conserv. Mar. Freshwat. Ecosyst. 12 (4), 365–379. Newson, M.D., Clark, M.J., in press. The sustainable management of restored rivers. In: Darby, S.E. and Sear, D.A. (Eds), Uncertainty in River Restoration, Wiley, Chichester, UK. NRC, 1999. New strategies for America’s watersheds: report of National Research Council. National Academy Press, Washington, DC. Pollack, H.N., 2003. Uncertain sciencey.uncertain world. Cambridge University Press, Cambridge, UK, 243pp. Reiser, D.W., 1998. Sediment in gravel bed rivers: ecological and biological considerations. In: Beschta, R.L., Komar, P.D., and Bradley, G. (Eds), Gravel-Bed Rivers in the Environment. Water Research Centre, Colorado, pp. 199–228. Richards, K., Brasington, J., Hughes, F., 2002. Geomorphic dynamics of ﬂoodplains: ecological implications and a potential modelling strategy. Freshwat. Biol. 47 (4), 559–579. Rotmans, J., Van Asselt, M.B.A., 2001. Uncertainty management in integrated assessment modeling: towards a pluralistic approach. Environ. Monit. Assess. 69 (2), 101–130. Rutherfurd, I., Montgomery, D.R., Hobbs, R., in press. Chapter ﬁve: uncertainty in designing restored river channels. In: Darby, S.E. and Sear, D. (Eds), Uncertainties in River Restoration. Wiley, Chichester, UK. Sear, D., 1996. The sediment system and channel stability. In: Brookes, A. and Shields, F.D. (Eds), River Channel Restoration: Guiding Principles for Sustainable Projects. Wiley, Chichester, UK, pp. 149–177. Sear, D.A., 1994. River restoration and geomorphology. Aquat. Conserv. Mar. Freshwat. Ecosyst. 4 (2), 169–177. Sear, D.A., Arnell, N., 2006. The application of paleohydrology in river management. Catena 66, 169–183. Sear, D., Darby, S.E., Scaife, R., 2001. Cherwell catchment restoration scenarios. Final Report for EU LIFE Project on Wise Use of Floodplains. Final Report, GeoData Institute, University of Southampton, Southampton, UK. Sear, D.A., Wilcock, D., Robinson, M.R., Fisher, K.R., 2000. Channel modiﬁcations and impacts. In: Acreman, M.C. (Ed.), The Changing Hydrology of the UK. Routledge, London, pp. 55–81. Van Asselt, M.B.A., 2000. Perspectives on uncertainty and risk: The PRIMA approach to decision support. Ph.D. Thesis, Kluwer Academics Publishers, Dordrecht, The Netherlands, 452pp. Van Asselt, M.B.A., Rotmans, J., 2002. Uncertainty in integrated assessment modeling – from positivism to pluralism. Clim. Chang. 54 (1–2), 75–105.

760

D.A. Sear, J.M. Wheaton, S.E. Darby

Walters, C.J., 1997. Challenges in adaptive management of riparian and coastal ecosystems. Conserv. Ecol. [online], 1(2): available at: http://www.consecol.org/vol1/iss2/art1 Ward, J.V., Tockner, K., Arscott, D.B., Claret, C., 2002. Riverine landscape diversity. Freshwat. Biol. 47 (4), 517–539. Wheaton, J.M., 2004. The signiﬁcance of ecohydraulic and geomorphic uncertainties in river restoration. Mini-thesis submitted in partial fulﬁllment for the transfer from Master of Philosophy (MPhil) to Doctor of Philosophy (Ph.D) Thesis, University of Southampton, Southampton, UK, 80pp. Wheaton, J.M., Darby, S.E., Sear, D., in press. Chapter three: scope of uncertainty in river restoration. In: Darby, S.E. and Sear, D. (Eds.), Uncertainties in River Restoration. Wiley, Chichester, UK. Wheaton, J.M., Milne, J.A., Darby, S.E., Sear, D.A., 2006. Does scientiﬁc conjecture accurately describe restoration practice? Insight from an International River Restoration Survey. Area 38 (2), 128–142. Wheaton, J.M., Pasternack, G.B., Merz, J.E., 2004. Spawning habitat rehabilitation – II. Using hypothesis testing and development in design, Mokelumne River, California, USA. Int. J. River Basin Manage. 2 (1), 21–37. Wilcock, P., 1997. Friction between river science and practice: the case of river restoration. EOS, Trans. Am. Geophys. Union, 78(40). Williams, P., Whitﬁeld, M., Biggs, J., Bray, S., Fox, G., Nicolet, P., Sear, D.A., 2004. Comparative biodiversity of rivers, streams, ditches and ponds in an agricultural landscape in Southern England. Biol. Conserv. 115 (2), 329–341. Wissmar, R.C., Beschta, R.L., 1998. Restoration and management of riparian ecosystems: a catchment perspective. Freshwat. Biol. 40, 571–585. Wissmar, R.C., Bisson, P.A. (Eds), 2003. Strategies for Restoring River Ecosystems: Sources of Variability and Uncertainty in Natural and Managed Systems. American Fisheries Society, Bethesda, MD, 270pp. Wissmar, R.C., Braatne, J.H., Beschta, R.L., Rood, S.B., 2003. 5. Variability of riparian ecosystems: implications for restoration. In: Wissmar, R.C., Bisson, P.A., and Duke, M. (Eds), Strategies for Restoring River Ecosystems: Sources of Variability and Uncertainty in Natural and Managed Systems. American Fisheries Society, Bethesda, MD, pp. 107–127. Wolman, M.G. and Leopold, L.B., 1957. River ﬂood plains: some observations on their formation. Geological Survey Professional Paper 282-C, United States Geological Survey, Washington, DC. WWF, 2001. The Status of Wild Atlantic Salmon: A River by River Assessment, World Wildlife Fund-US, Washington; World Wildilfe Fund-Norway, Oslo; World Wildlife Fund – European Freshwater Programme, Copenhagen, Denmark.

Discussion by Gary Williams I very much agree with the need to consider uncertainty and present information about risks. Life is always uncertain and people do understand this. We should be presenting information in a way that makes the risks known. In many ﬁelds, not just in the gravelbed rivers fraternity, people are grappling with risk, and developing ways of analysing and presenting information on risk. I see this paper as part of that process of risk analysis and the development of formal procedures for dealing with uncertainties.

Discussion by Gordon Grant With river restoration emerging as an international focus of both river management agencies and scientists, it is interesting to note the different disciplinary ‘‘cultures’’ driving restoration in different parts of the world. One can view this perhaps as a ternary diagram with the three major disciplines involved – engineering, ecology and

760

D.A. Sear, J.M. Wheaton, S.E. Darby

Walters, C.J., 1997. Challenges in adaptive management of riparian and coastal ecosystems. Conserv. Ecol. [online], 1(2): available at: http://www.consecol.org/vol1/iss2/art1 Ward, J.V., Tockner, K., Arscott, D.B., Claret, C., 2002. Riverine landscape diversity. Freshwat. Biol. 47 (4), 517–539. Wheaton, J.M., 2004. The signiﬁcance of ecohydraulic and geomorphic uncertainties in river restoration. Mini-thesis submitted in partial fulﬁllment for the transfer from Master of Philosophy (MPhil) to Doctor of Philosophy (Ph.D) Thesis, University of Southampton, Southampton, UK, 80pp. Wheaton, J.M., Darby, S.E., Sear, D., in press. Chapter three: scope of uncertainty in river restoration. In: Darby, S.E. and Sear, D. (Eds.), Uncertainties in River Restoration. Wiley, Chichester, UK. Wheaton, J.M., Milne, J.A., Darby, S.E., Sear, D.A., 2006. Does scientiﬁc conjecture accurately describe restoration practice? Insight from an International River Restoration Survey. Area 38 (2), 128–142. Wheaton, J.M., Pasternack, G.B., Merz, J.E., 2004. Spawning habitat rehabilitation – II. Using hypothesis testing and development in design, Mokelumne River, California, USA. Int. J. River Basin Manage. 2 (1), 21–37. Wilcock, P., 1997. Friction between river science and practice: the case of river restoration. EOS, Trans. Am. Geophys. Union, 78(40). Williams, P., Whitﬁeld, M., Biggs, J., Bray, S., Fox, G., Nicolet, P., Sear, D.A., 2004. Comparative biodiversity of rivers, streams, ditches and ponds in an agricultural landscape in Southern England. Biol. Conserv. 115 (2), 329–341. Wissmar, R.C., Beschta, R.L., 1998. Restoration and management of riparian ecosystems: a catchment perspective. Freshwat. Biol. 40, 571–585. Wissmar, R.C., Bisson, P.A. (Eds), 2003. Strategies for Restoring River Ecosystems: Sources of Variability and Uncertainty in Natural and Managed Systems. American Fisheries Society, Bethesda, MD, 270pp. Wissmar, R.C., Braatne, J.H., Beschta, R.L., Rood, S.B., 2003. 5. Variability of riparian ecosystems: implications for restoration. In: Wissmar, R.C., Bisson, P.A., and Duke, M. (Eds), Strategies for Restoring River Ecosystems: Sources of Variability and Uncertainty in Natural and Managed Systems. American Fisheries Society, Bethesda, MD, pp. 107–127. Wolman, M.G. and Leopold, L.B., 1957. River ﬂood plains: some observations on their formation. Geological Survey Professional Paper 282-C, United States Geological Survey, Washington, DC. WWF, 2001. The Status of Wild Atlantic Salmon: A River by River Assessment, World Wildlife Fund-US, Washington; World Wildilfe Fund-Norway, Oslo; World Wildlife Fund – European Freshwater Programme, Copenhagen, Denmark.

Discussion by Gary Williams I very much agree with the need to consider uncertainty and present information about risks. Life is always uncertain and people do understand this. We should be presenting information in a way that makes the risks known. In many ﬁelds, not just in the gravelbed rivers fraternity, people are grappling with risk, and developing ways of analysing and presenting information on risk. I see this paper as part of that process of risk analysis and the development of formal procedures for dealing with uncertainties.

Discussion by Gordon Grant With river restoration emerging as an international focus of both river management agencies and scientists, it is interesting to note the different disciplinary ‘‘cultures’’ driving restoration in different parts of the world. One can view this perhaps as a ternary diagram with the three major disciplines involved – engineering, ecology and

760

D.A. Sear, J.M. Wheaton, S.E. Darby

Walters, C.J., 1997. Challenges in adaptive management of riparian and coastal ecosystems. Conserv. Ecol. [online], 1(2): available at: http://www.consecol.org/vol1/iss2/art1 Ward, J.V., Tockner, K., Arscott, D.B., Claret, C., 2002. Riverine landscape diversity. Freshwat. Biol. 47 (4), 517–539. Wheaton, J.M., 2004. The signiﬁcance of ecohydraulic and geomorphic uncertainties in river restoration. Mini-thesis submitted in partial fulﬁllment for the transfer from Master of Philosophy (MPhil) to Doctor of Philosophy (Ph.D) Thesis, University of Southampton, Southampton, UK, 80pp. Wheaton, J.M., Darby, S.E., Sear, D., in press. Chapter three: scope of uncertainty in river restoration. In: Darby, S.E. and Sear, D. (Eds.), Uncertainties in River Restoration. Wiley, Chichester, UK. Wheaton, J.M., Milne, J.A., Darby, S.E., Sear, D.A., 2006. Does scientiﬁc conjecture accurately describe restoration practice? Insight from an International River Restoration Survey. Area 38 (2), 128–142. Wheaton, J.M., Pasternack, G.B., Merz, J.E., 2004. Spawning habitat rehabilitation – II. Using hypothesis testing and development in design, Mokelumne River, California, USA. Int. J. River Basin Manage. 2 (1), 21–37. Wilcock, P., 1997. Friction between river science and practice: the case of river restoration. EOS, Trans. Am. Geophys. Union, 78(40). Williams, P., Whitﬁeld, M., Biggs, J., Bray, S., Fox, G., Nicolet, P., Sear, D.A., 2004. Comparative biodiversity of rivers, streams, ditches and ponds in an agricultural landscape in Southern England. Biol. Conserv. 115 (2), 329–341. Wissmar, R.C., Beschta, R.L., 1998. Restoration and management of riparian ecosystems: a catchment perspective. Freshwat. Biol. 40, 571–585. Wissmar, R.C., Bisson, P.A. (Eds), 2003. Strategies for Restoring River Ecosystems: Sources of Variability and Uncertainty in Natural and Managed Systems. American Fisheries Society, Bethesda, MD, 270pp. Wissmar, R.C., Braatne, J.H., Beschta, R.L., Rood, S.B., 2003. 5. Variability of riparian ecosystems: implications for restoration. In: Wissmar, R.C., Bisson, P.A., and Duke, M. (Eds), Strategies for Restoring River Ecosystems: Sources of Variability and Uncertainty in Natural and Managed Systems. American Fisheries Society, Bethesda, MD, pp. 107–127. Wolman, M.G. and Leopold, L.B., 1957. River ﬂood plains: some observations on their formation. Geological Survey Professional Paper 282-C, United States Geological Survey, Washington, DC. WWF, 2001. The Status of Wild Atlantic Salmon: A River by River Assessment, World Wildlife Fund-US, Washington; World Wildilfe Fund-Norway, Oslo; World Wildlife Fund – European Freshwater Programme, Copenhagen, Denmark.

Discussion by Gary Williams I very much agree with the need to consider uncertainty and present information about risks. Life is always uncertain and people do understand this. We should be presenting information in a way that makes the risks known. In many ﬁelds, not just in the gravelbed rivers fraternity, people are grappling with risk, and developing ways of analysing and presenting information on risk. I see this paper as part of that process of risk analysis and the development of formal procedures for dealing with uncertainties.

Discussion by Gordon Grant With river restoration emerging as an international focus of both river management agencies and scientists, it is interesting to note the different disciplinary ‘‘cultures’’ driving restoration in different parts of the world. One can view this perhaps as a ternary diagram with the three major disciplines involved – engineering, ecology and

Uncertain restoration of gravel-bed rivers and the role of geomorphology

761

geomorphology – represented on the vertices and the approaches used in various countries plotted in terms of the relative contribution of each discipline to the restoration enterprise (Fig. 28.7). In the US, restoration is primarily motivated and implemented by ecologists, with lesser contributions from geoscientists and engineers. In Japan, engineering approaches are dominant, with lesser contributions from the ecological disciplines. Europe seems somewhat in between these two. It will be interesting to see how these different disciplinary mixes result in different onthe-ground strategies, and to follow up on the successes or failures of each. All of this must be inset, of course, within the larger social and cultural context of what ‘‘restoration’’ means for each society, which will deﬁne both the goals and the basis for evaluating effectiveness. Ecology

U.S.

Europe Japan

Engineering

Geomorphology

Figure 28.7. Ternary diagram showing disciplinary position of national restoration approaches.

Reply by the authors The authors thank the two discussers for raising some interesting points. The notion of different weights between the disciplinary components based on geographical traditions is interesting, but changes with time. In the UK, for example the main driver for restoration is now conservation legislation rather than ﬂood risk management. The optimum response must surely come from a balanced and appropriate combination of all disciplines – again this tends to be project speciﬁc.

Uncertain restoration of gravel-bed rivers and the role of geomorphology

761

geomorphology – represented on the vertices and the approaches used in various countries plotted in terms of the relative contribution of each discipline to the restoration enterprise (Fig. 28.7). In the US, restoration is primarily motivated and implemented by ecologists, with lesser contributions from geoscientists and engineers. In Japan, engineering approaches are dominant, with lesser contributions from the ecological disciplines. Europe seems somewhat in between these two. It will be interesting to see how these different disciplinary mixes result in different onthe-ground strategies, and to follow up on the successes or failures of each. All of this must be inset, of course, within the larger social and cultural context of what ‘‘restoration’’ means for each society, which will deﬁne both the goals and the basis for evaluating effectiveness. Ecology

U.S.

Europe Japan

Engineering

Geomorphology

Figure 28.7. Ternary diagram showing disciplinary position of national restoration approaches.

Reply by the authors The authors thank the two discussers for raising some interesting points. The notion of different weights between the disciplinary components based on geographical traditions is interesting, but changes with time. In the UK, for example the main driver for restoration is now conservation legislation rather than ﬂood risk management. The optimum response must surely come from a balanced and appropriate combination of all disciplines – again this tends to be project speciﬁc.

Gravel-Bed Rivers VI: From Process Understanding to River Restoration H. Habersack, H. Pie´gay, M. Rinaldi, Editors r 2008 Elsevier B.V. All rights reserved.

763

29 Historical channel modiﬁcation and ﬂoodplain forest decline: implications for conservation and restoration of a large ﬂoodplain river – Willamette River, Oregon Stanley Gregory

Abstract Trajectories of change in channel structure and riparian plant communities have been documented for the 273-km mainstem of the Willamette River from Eugene to Portland, OR, USA. We also map current human systems (population density, buildings and roads, public lands, land values, land use) as measures of social opportunities and constraints. We use this channel-change detection and human systems analysis as a basis for spatially explicit prioritization of potential restoration efforts. Priorities for conservation of relatively functional reaches are based on current conditions of the channel and ﬂoodplain forest along the river. We also measured the consequences in future alternatives as described by stakeholders in the Willamette River basin. Scenarios of change from 2000 to 2050 were developed for current policies and practices, development alternatives, and conservation options. We compare patterns of recent ﬂoods to historical channels to provide estimates of the potential for natural ﬂood processes to restore biophysical structure and function in ﬂoodplain rivers. These quantitative evaluations of historical changes and future trajectories of ecological properties of the Willamette River will be used to identify potential strategies for restoration of a large river during a period of rapid population growth. 1.

Introduction

Human history is written in changing landscapes – forests converted to villages and ﬁelds, grasslands converted to agriculture, villages coalescing into cities, rivers straightened and contained (Petts, 1990; Hulse and Gregory, 2001). The societies that create such change also depend on these same landscapes for natural resources and a livable environment. Use of certain resources (e.g., gravel, trees, water) leads to the loss of other resources (e.g., ﬁsh, macroinvertebrates, amphibians, birds, mammals). E-mail address: [email protected] ISSN: 0928-2025

DOI: 10.1016/S0928-2025(07)11163-9

764

S. Gregory

As rivers ﬂow by communities, resource managers, politicians, and individual citizens must look to a wide array of specialists for information and advice on their future options for creating a landscape in which they can meet their needs for food, water, land, commerce, transportation, and recreation. Inevitably, competing needs lead to challenging decisions about conserving existing landscape features and restoring landscape elements lost through either natural processes or past human decisions (Petts, 1990; Gregory et al., 1998; Baker et al., 2004). This paper describes historical changes in channel geomorphology in the Willamette River, a large anastomosing river in the Paciﬁc Northwest region of the United States (Gregory et al., 2002a,d; Hulse et al., 2002). The history of channel change and loss of islands will be used to illustrate potential quantitative approaches for designing and communicating opportunities for conservation and restoration of a large river. The management of ﬂoodplain rivers commonly focuses on navigation, ﬂood control, water consumption, protection of adjacent property, and channel conﬁnement (Naiman and Decamps, 1990; IFMRC, 1994). Geomorphologists, civil engineers, hydrologists, ecologists, economists, rural land owners, urban land owners, and regional planners often ﬁnd themselves either representing conﬂicting interests in decisions that are largely irreversible or being left out of the decision-making process all together (Hyman and Leibowitz, 2000; Baker et al., 2004). Our existing landscapes are mosaics of the outcomes of innumerable past actions by individuals, communities, regional and national government agencies, most with little attention to the overall form and function of a river ecosystem. One of the most common causes of river simpliﬁcation is the loss of side channels and islands. Several processes can convert reaches with multiple channels to single channels (Petts and Foster, 1985; Gregory, 1992; Gurnell and Petts, 1995). Islands composed of alluvium can be eliminated, either through natural processes of channel erosion or through human removal. Gravel mining is a common practice that reduces or eliminates alluvial islands, but dredging for navigation or ﬂow modiﬁcation also can cause sediment scour and loss of bars and islands. In this case, the land that once existed as an island no longer exists. A second major process that eliminates islands in rivers is the closing of side channels and hardening of the banks in the vicinity of the channel closure. In this case, most of the area of the island remains but now functions as a river bank contiguous with the adjacent terrestrial ecosystem. Both mechanisms of island loss decrease the hydraulic and geomorphic complexity of the river, but they have much different outcomes in terms of area of ﬂoodplain and associated riparian habitat. The major processes of island loss change the hydraulic resistance to ﬂow and geomorphic bedforms (Schumm, 1968; Gurnell and Petts, 1995). The modiﬁcations of the river channel have important consequences for ecological processes and aquatic ecosystems (Naiman et al., 1988; NRC, 1992; Van Sickle et al., 2004). Changes in habitat structure alter community composition and population abundance of plants, invertebrates, ﬁsh, and other vertebrates. These alterations also change ﬂoodplain plant communities and the riparian processes that inﬂuence aquatic ecosystems. Changes in hydraulics and bedform also modify hyporheic exchange, the interaction of surface water and subsurface ﬂow (Gregory and Gurnell, 1988; Swanson et al., 1998). This may alter the thermal heterogeneity of the river.

Historical channel modification and floodplain forest decline

765

Examples of biological and thermal patterns in the Willamette River illustrate these potential effects of channel modiﬁcation. The Willamette River ﬂows north across the Willamette Valley for 273 km before entering the larger Columbia river close to Portland (Gregory et al., 2002a). Local landforms, such as volcanic cones, ridges, deposits of glacial outburst ﬂoods, block and conﬁne the river along its length. The ﬂat topography of the valley ﬂoor was caused by sediment deposition from a series of ﬂoods that inundated much of the Willamette Valley ﬂoor during the end of the last glacial period. The upstream portion of the Willamette Valley Eugene to Albany contains an anastomosing channel (Fig. 29.1). The channel in this upper portion of the mainstem remains the most complex reach of the river, but it has been simpliﬁed in the last 150 years through channel modiﬁcation. Eight major tributaries (Middle Fork Willamette River, Coast Fork Willamette River, McKenzie River, Long Tom River, Marys River, Calapooia River, Santiam River, and Luckiamute River) have transported alluvial sediment into a depositional basin created by the blockages of the Salem hills. The low gradient and extensive alluvial deposits result in the anastomosing pattern of this reach of the river. The middle reach of the Willamette River extends from Albany to Newberg. Several large hills and ridgelines from the adjacent Coast Range to the west and Cascade Range to the east exert varying inﬂuences on the channel along this reach. As a result, the channel form exhibits a mixture of anastomosing channels and single thread channels. The downstream end of the Willamette River is a simple meandering river channel with lower gradients than the upper river section. The Willamette Falls exert a major control on the river, and the 45 miles (72 km) below Willamette Falls is extremely low gradient and controlled by the backwatering effect of the Columbia River. Complex braided channels are more localized, and lateral changes in the river channel are limited. Most islands in this reach are volcanic outcrops with far less alluvial deposits as in the islands and bars of upper reaches. In this study, we assess the channel change and loss of islands in the Willamette River from 1850 to 2000 and relate differences in channel dynamics to the geology of the basin and spatial patterns of human modiﬁcation. Transitions from island to lateral ﬂoodplain are identiﬁed and implications for ecological processes are discussed.

2.

Methods

Channel change from 1850 to 2000 was measured for the 273-km mainstem of the Willamette River from Eugene, OR to its conﬂuence with the Columbia River in Portland, OR. The Willamette River and its major tributaries were surveyed by the U.S. General Land Ofﬁce (GLO) in the decade after 1850. Engineers surveyed in a grid-based system using a longitudinal axis along the west coast of the United States (the Willamette meridian). The surveyors delineated changes in vegetation, stream channels, and wetlands along monumented section lines and sketched the features within a one-square mile section of land. The Nature Conservancy transcribed the engineering coordinates and notes and we transferred the maps into a GIS database

766

Figure 29.1. Map of the 30,000-km2 Willamette River basin, Oregon, USA.

S. Gregory

Historical channel modification and floodplain forest decline

767

(1:24,000 scale). The Willamette River network and its riparian vegetation also were mapped in 1895 and 1932 by the U.S. Army Corps of Engineers. These two river surveys were conducted for the entire length of the Willamette River for navigation purposes. We used satellite imagery (Landsat) and digital orthophotographs to construct a map of the Willamette River in 1995 (recently updated for 2000). The river channels were classiﬁed as primary channel (unbraided, or the portion of the channel with the most ﬂow), side channel (channels connected to the mainstem at both ends; this includes the small channels that form islands), and alcoves or sloughs. The latter are ‘‘blind’’ channels, i.e., they are connected only at one end to a primary, side, or secondary channel. Islands and gravel bars were identiﬁed from these surveys and represented in the GIS layers for each year. Channel change was quantiﬁed by estimating both length and area of the major channel types between years. This paper will focus on the overall change in the river channel and islands from 1850 to 1995. Change in island extent was measured ﬁrst as an overall change in area of islands. This measure of island change cannot differentiate between sources of island loss (i.e., elimination of the area of the island or the elimination of a side channel converting the island into the terrestrial margin or riverbank). To distinguish these two processes, we visually examined each island along the 273-km length of the river and classiﬁed it into ﬁve major classes: (i) unchanged – original island largely the same as 1850, (ii) remnant – portion of the original island still existing, (iii) lost from remnant – portion of remnant that was lost, (iv) extinct – island no longer present and original area occupied by river, and (v) transformed – island no longer present but land area now adjacent to river. The data were converted to percentage of the original area of islands in 1850. No new islands were observed. Results are presented for the entire river and the three major geomorphic reaches. Thermal patterns of the Willamette River currently are being investigated and we present an example from a recent study to illustrate the consequence of island and channel change on the thermal heterogeneity of a large river. Thermal data loggers were deployed for 1 week at more than 50 locations in an island reach of the upper Willamette in late July 2005. Daily maxima for all stations were determined and used to identify areas of likely exchange of water from the hyporheos to the river.

3. 3.1.

Results Changes in channel and island area

From 1850 to 1995, total area of river channels decreased by 22% (Table 29.1, Gregory et al., 2002a,b) and total length of all channels decreased by 26%. More than 30% of alcoves and sloughs were lost by 1995. Islands diminished more than any other channel type. Total area of islands decreased by 63% over the 145-year interval. These changes differed along the length of the Willamette River. The upstream reach from Albany to Eugene experienced the greatest change in channel structure and loss of islands. This reach ﬂows through broad alluvial and

S. Gregory

768

Table 29.1. Area of channel types and islands in 1850 and 1995 reported in hectares for the three reaches of the Willamette River and the combined length of the mainstem river. River section 1850 Upper Middle Lower Total mainstem 1995 Upper Middle Lower Total mainstem Percent change Upper Middle Lower Total mainstem

Primary channel

Secondary channel

Alcoves

Islands

Total area of channel

1946 2406 1473 5825

1059 309 110 1478

181 81 10 272

6897 1946 122 8965

10,083 4742 1715 16,540

1533 2115 1406 5054

280 209 169 658

103 71 2 176

1413 1777 117 3307

3329 4172 1694 9195

–73.5 –33.1 53.6 –55.6

–41.5 –15.5 –80.0 –35.1

–21.3 –12.1 –4.6 –13.3

–79.5 –8.6 –3.0 –63.1

–39.8 –14.5 –1.1 –22.3

Source: Modiﬁed from Gregory et al. (2002). Note: Percent change from 1850 to 1995 is reported following the areas for 1850 and 1995 (negative values for percent change represent loss of area). Total area of channel includes areas of the wetted channel types plus area of islands.

glacial ﬂood deposits. During the period from 1850 to 1995, total area of river channels and islands (combined) in this reach decreased by 40% (Table 29.1). The total length of all channels decreased from 340 to 185 km (Gregory et al., 2002a; http://oregonstate.edu/dept/pnw-erc). More than 70% of the side channels and 40% of the alcoves were lost. Approximately 80% of the islands in this reach have been eliminated or converted to ﬂoodplain banks. The middle reach from Albany to Newberg exhibited variable patterns of channel change and was intermediate in loss of channels and islands. Basaltic outcrops and ridges from the foothills caused local constraint and channel control in the middle reach. The channel of the Willamette River in this reach has been simpliﬁed, though not to the extent observed in the upper river. Total area of channels and islands combined decreased by approximately 15%. Most of that change occurred through loss of primary channels and islands, though the proportional change was greatest for side channels (Table 29.1). Total length of channels did not change. The lower reach from Newberg to the mouth in Portland changed very little from 1850 to 1995. Lava ﬂows from central Oregon created a basaltic trench in the lower reach more than 10 million years ago and make it less vulnerable to natural and human alteration. Though proportional changes were substantial, changes in area of major channel features were minor. Total area of channels and islands combined decreased by 22%, but even in 1850 the area of these geomorphic features was substantially less than the upper two reaches. Total length of channels increased from 69 to 76 km.

Historical channel modification and floodplain forest decline 3.2.

769

Loss of Islands

Our analysis of islands in the Willamette River in 1995 revealed that very few islands from 1850 remain intact and unchanged today (Table 29.2). For the total Willamette River mainstem, only 1% of the area of islands remains unchanged. One-third of the island area now exists as a remnant of an island of 1850. The portion of those remnant islands that has been lost accounts for 14% of the island area in 1850. Almost half of the islands have become the banks of the river along the ﬂoodplain. However, only 5% of the islands of 1850 have been totally eliminated. This pattern of island change differs greatly for the three major reaches of the Willamette River (Table 29.2). In the upper river, where most of the overall change in islands has been observed, most of the island area has become riverbank. None of the original islands remain unchanged. Remnant islands account for almost 20% of the 1850 island extent, and almost 25% of the island area has been truly lost (extinct plus loss from remnant islands). In sharp contrast, the middle reach of the river predominantly (84%) exists as remnants of the islands of 1850. Only 11% of the 1850 island extent has been converted to ﬂoodplain riverbank. Less than 2% of the area of islands in 1850 has been truly lost in the middle reach. The lower reach, where relatively little channel change occurred, exhibits a totally different pattern, with a substantial proportional creation of new islands (16% of the area of 1850 islands). Over the same interval, 44% of the 1850 island area has been totally eliminated and another 13% has been lost from remnant islands. These patterns of island change indicate that analysis of channel change requires closer inspection than simple change in island extent. Human activity and natural geomorphic processes can cause markedly different processes of island alteration to occur. In the Willamette River, most of the change in islands did not occur in the areas of greatest human density, though human activities were responsible for much of the observed change. Portland and Salem (located in the lower and middle reaches, respectively) are the largest cities along the Willamette River. These reaches exhibited only 3 and 9% loss of island area since 1850. The upper reach has been shaped by depositions of sediments from glacial outburst ﬂoods 15,000–20,000 years ago and subsequent deposition of alluvial sediment. This section of the river contained the greatest length of anastomosing channels and associated islands. The major land use in this reach of river is agriculture with small cities distributed along its length. Secondary channels have been blocked, some levees have been built, and Table 29.2.

Sources of change in island extent from 1850 to 1995.

Reach

New

Extinct Transformed

Unchanged

Remnant

Loss from remnant

Upper Middle Lower Total mainstem

1.1 4.4 15.5 2.1

5.3 1.2 43.7 5.3

0.0 3.6 8.3 0.9

19.2 84.0 35.5 33.3

18.3 0.4 12.5 14.3

57.3 10.7 0.0 46.1

Note: Changes are expressed as percent of the total area of islands in 1850.

S. Gregory

770

numerous banks have been hardened with riprap. We mapped the locations where banks had been hardened with boulders and found that only 26% of the length of the mainstem river had been hardened. However, these structures have been placed in the most dynamic locations along the channel, occurring on 67% of the major meanders in the river (Gregory et al., 2002c). The geomorphic structure of the river in both 1850 and 1995 largely reﬂects the template created by the ﬂuvial geomorphology and local inﬂuences of valley landforms in the different reaches. Humans have greatly altered the geomorphology of the Willamette River but not primarily in the areas of highest population density of human populations.

3.3.

Prioritization of future conservation and restoration actions

In 1850, the 273-km distance along the mainstem Willamette River from Eugene, OR downstream to Portland, OR contained 571 km of channels (e.g., side channel, sloughs, or alcoves) (Gregory et al., 2002a). By 1995, the length of channels over this distance had decreased to 424 km, a 25% loss of channel complexity. The ﬂoodplain forest along the river was changed even more by land conversion. In 1850, more than 4360 ha of ﬂoodplain forest surrounded the 273-km distance along the mainstem. By 1990, the extent of ﬂoodplain forest had been reduced to 1197 ha, a loss of 73% of the historical ﬂoodplain forest area. These changes in the river and its ﬂoodplain potentially alter the abundance and distribution of ﬁsh communities in the mainstem river. Our ﬁeld studies in the Willamette River have demonstrated that ﬁsh abundance is approximately 50% greater in reaches with multiple channels or tributary junctions. In addition, we ﬁnd approximately 20 ﬁsh species in a 1-km reach of the Willamette River that contains multiple channels or tributary junctions, but we ﬁnd less than 17 species in the single-thread channels. We incorporated these relationships into an analysis of potential for future restoration in the Willamette River. We created a spatial framework for the assessment of the river and its ﬂoodplain by dividing the area of river that has been inundated between 1850 and the present into 1-km slices along the central axis of the channel (Fig. 29.2). We evaluated the biophysical potential for increased channel complexity, ﬂoodplain forest, and ﬂood storage (through removal of revetments) based on the difference between 1850 and 1995 for each 1-km slice (Hulse and Gregory, 2004). We identiﬁed areas of socioeconomic obstacles for the same areas based on population density, buildings and roads, property value, and public land. Based on these two constraints on restoration success, we created a template to illustrate areas of higher potential for ecological restoration along the Willamette River (Fig. 29.3). This framework is now being used by the Willamette Partnership and other agencies and citizens groups to guide conservation and restoration efforts in the Willamette Valley.

3.4.

Consequences for thermal heterogeneity in river water

Channel avulsion and ﬂoodplain coalescence can alter the extent, position, and function of islands (Pie´gay and Bravard, 1997; Pie´gay, 1998). The distinction between

Historical channel modification and floodplain forest decline

771

slice 0

Portland Metro

Oregon

slice 50 slice 150 Albany slice 100

Salem/Keizer

Corvallis

Willamette River Basin Albany

Peoria

Corvallis

slice 200

slice 200

Harrisburg

Eugene/ Springfield

Figure 29.2. Longitudinal framework for analysis of conditions within the historical ﬂoodplain in 1-km ‘‘slices’’ perpendicular to ﬂoodplain axis. For scale, distances between lines along the ﬂoodplain axis are 1 km and numbering begins near the conﬂuence of the Willamette River and the Columbia River and extends 227 km upstream to above the conﬂuence with the McKenzie River. These 1-km slices were used for prioritization of locations for river restoration. (Reproduced with permission from Hulse and Gregory, 2004, Springer.)

true loss of island area and transformation of islands to riverbanks has important geomorphic and ecological implications. To explore the impact of physical heterogeneity on thermal patterns in the river water, we placed temperature data loggers around island features in the upper Willamette River. Most of the habitats associated with the primary channel exhibited maximum temperatures ranging from 19.1 to 19.81C. Temperature did not differ greatly with depth (o11C), indicating turbulent ﬂow and thorough mixing. However, several springs were observed emerging from the ﬂoodplain and gravel bar that exhibited maximum temperatures of 14.6, 15.4, and 15.81C. We investigated ﬁve additional reaches during 2005 and all reaches contained cold water habitats with temperatures ranging from 2.0 to 8.01C colder than the mainstem. These colder springs most likely represent hyporheic exchange with the surface water and are created by ﬂow through the gravel bar and transport through the alluvium of islands and gravel bars. In the Willamette River, such subsurface exchange provides important cold water refuges for coldwater species, such as Cutthroat trout (Oncorhynchus clarki clarki), Chinook salmon (Oncorhynchus tshawytscha), and Steelhead (Oncorhynchus mykiss). Loss of island features is likely to reduce the frequency and thermal inﬂuence of the subsurface

S. Gregory

772 0 10 20 30 40

80

70

60

50

90 100 110 120 130 140

160

150

170

190 200 210

Demographic and Economic Constraints Index

1.00

180

Slice 198 0.50

Slice 190 0.00

0.00

Slice 189

0.50 Biophysical Potential Index

1.00

220

Figure 29.3. Illustration of high priority sites for restoration based on potential improvement on (1) increased channel complexity, (2) increased area of ﬂoodplain forest, (3) increased non-structural ﬂood storage. (Reproduced with permission from Hulse and Gregory, 2004, Springer.) Dark gray bands indicate areas with high ecological potential for restoration and low socioeconomic obstacles. White bands are areas with high ecological potential for restoration and high socioeconomic obstacles. Light gray bands indicate areas with low ecological potential for restoration and low socioeconomic obstacles, and medium gray bands are areas with low ecological potential for restoration and high socioeconomic obstacles. For scale, distances between lines along the ﬂoodplain axis are 1 km.

Historical channel modification and floodplain forest decline

773

exchange and lead to thermal simpliﬁcation and warmer overall river temperatures (Gregory and Bisson, 1996).

4.

Discussion and conclusions

Human modiﬁcation of the river and its basin has simpliﬁed the Willamette River from 1850 to the present. Land owners and agencies directly blocked side channels with boulders, wood pilings, and ﬁll in an attempt to prevent channel change and loss of property along the river. In addition, 26% of the length of the river has been hardened by lining the bank with boulders (riprap) on either one bank or both banks. In addition, gravel has been mined from the river since the late 1800s and aggregate industries continue to mine gravel bars in the main channel. Straightening the channel, eliminating side channels, and removing gravel increase channel degradation further isolate the active channel from its ﬂoodplain and accelerates the loss of side channels and islands. Klingeman (1973) analyzed changes in the relationship between staff gage elevation and discharge at gaging stations along the Willamette River. He concluded that the mainstem Willamette River was changing rapidly and was not in a dynamic equilibrium hydrologically. Most sites exhibited streambed degradation, though lateral channel adjustment was noted at some sites. The rate of channel degradation or downcutting for the mainstem river was estimated to be approximately 0.3 m/decade. Klingeman (1973) cited upstream dams, gravel mining in the river, and streambank hardening (i.e., riprap) as likely mechanisms responsible for the observed channel degradation, which is consistent with our observations of channel change in the Willamette River since 1850. Two additional factors – changes in discharge and sediment delivery to the mainstem river – are potential mechanisms for channel simpliﬁcation in the Willamette River over the last 150 years. Flood-control dams alter the frequency of ﬂoods that exceed the bankfull channel (Magilligan et al., 2003) and reduce sediment delivery into downstream rivers (Ligon et al., 1995). Both of these processes could contribute to channel incision and loss of lateral channel complexity in the Willamette River. Flood management currently reduces the peak discharges and maintains high ﬂows at close to bankfull discharge but less than ﬂoodplain inundation discharges. This focuses the stream power within the active channel and may lead to channel incision. Likewise, reduction of sediment input into the mainstem river reduces the gravel load and causes continued erosion of older gravels deposits within the mainstem river. While dams may currently contribute to channel simpliﬁcation, they could only account for a minor portion of historical channel change because they were not built until after 1945. Channel simpliﬁcation in the Willamette River was clearly evident by 1895 and 1932, thus human alteration or other mechanisms were inﬂuencing the channel morphology prior to ﬂood control dams. One ﬁnal factor that could inﬂuence the changes in the Willamette River is regional climate change. Peak discharges in the Willamette River have declined since the late 1800s (Fig. 29.4), but mean annual discharge has remained relatively constant. Flood-control reservoir were built in the basin after 1945, therefore the decline in peak ﬂows between 1860 and 1945 cannot be attributed to dams. After 1965, ﬂood

S. Gregory

774 7500

Mean Daily Discharge (cms)

6500

5500

4500

3500

2500

1500 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 Year Figure 29.4. Daily mean discharges during ﬂoods greater than 1982 cms at the Albany, OR gaging station from 1895 to 2006 (data from USGS gaging station 14174000). The estimated 2-year recurrence discharge under ﬂow regulation is 1982 cms; the 2-year unregulated ﬂow is estimated to be 3200 cms. (Data from Portland, OR Ofﬁce of USACE). All daily mean discharges that exceeded the 2-year regulated discharge are plotted.

control reservoirs reduce most peak ﬂows in the mainstem Willamette River by 30–40%. Flow regulation has decreased 2-year recurrence ﬂows at the Albany gaging station from 3200 to 1980 cms (hydrologic record from 1895 to 2006). Reduced peak ﬂows over the last 150 years could increase the potential for incision and channel simpliﬁcation (as previously discussed). One of the major tools in ﬂoodplain restoration is the re-establishment of hydrologic and geomorphic processes to the extent possible and in river reaches where the outcomes are likely and the ecological beneﬁts are signiﬁcant (NRC, 1992; Hulse and Gregory, 2004). Restoration of complex braiding or anastomosing channels can be accomplished by reducing lateral human-created constraints along the banks, increasing delivery of sediments, restoring hydrologic regimes, and reconnecting historical channels. The degree of success will depend on the local hydraulic and geomorphic processes and the degree to which human modiﬁcations can be reversed or reduced. There is a tendency to consider human manipulations in restoring river complexity, but natural ﬂood processes may achieve such restoration more consistently and effectively than costly and time-consuming engineering approaches. Passive restoration through natural hydrologic and geomorphic processes obviously requires the reduction of human constraints, but it also requires the acceptance of dynamic and relatively unplanned and undirected channel change. This latter requirement commonly is the most difﬁcult to achieve. Scientists who study gravel-bed rivers appreciate the

Historical channel modification and floodplain forest decline

775

efﬁciency and effectiveness of using natural geomorphic and ecological processes in river restoration, but incorporation of these approaches into river management will require new and more effective methods of illustration and communication between collaboration of river scientists, river managers, and regional communities. One possible interpretation of the ﬁndings of this study could be that the total area of ﬂoodplain has not been greatly diminished even though 63% of the total area islands has been lost. Many of the islands became ﬂoodplains adjacent to the mainstem river, and the secondary channels were blocked and ﬁlled. Total elimination of islands occurred primarily in the lower, simpler reach of the river near Portland (44% became extinct). In the more complex upper reach of the river, 57% of the islands in 1850 were transformed to current adjacent ﬂoodplains. These areas where islands became ﬂoodplains still support many potential ﬂoodplain functions (e.g., ﬂood detention, soil deposition, wildlife habitat, riparian forests), though land use has converted many of these ﬂoodplains to crops, residences, or urban areas and altered the potential ﬂoodplain functions. Ecological functions of complex anastomosing or braided river channels extend far beyond the functions of islands as areas of ﬂoodplain. These channels provide extensive edge habitat with variable depths and velocities, secondary channels with greater roughness and reduced velocity, numerous gravel bars, alcoves, and sloughs. Such areas exhibit greater hyporheic exchange and provide mosaics of cold and warm water habitats. In addition, the greater roughness of such complex channels dissipates energy during ﬂoods and provides critical refuges for aquatic communities. Loss of channel complexity, secondary channels and islands potentially diminishes the species richness and population abundances of aquatic organisms in large rivers. Restoration of such complexity cannot be achieved simply by carving a few alcoves or side channels in an artiﬁcially hardened river. Restoration of river complexity requires a broader perspective of the river in which it is allowed to meander within its ﬂoodplain, erode and deposit alluvial sediments, and recruit wood and sediment dynamically from adjacent forests during ﬂoods. Communities must minimize channel hardening where possible and avoid development of costly buildings and roads in the most dynamic portions of the landscape. This can be best accomplished by protecting existing ﬂoodplain functions to the greatest extent possible and taking advantage of river changes that occur to major ﬂood events to back away from the river and allow it to restore itself through geomorphic processes.

Acknowledgements This work was funded by STAR grant R825797 between the U.S. Environmental Protection Agency and Oregon State University. Data on historical river changes and land use/land cover were developed as part of cooperative agreement CR824682 between USEPA and OSU. Supplementary information and the digital datasets referenced herein are available via the Internet at http://www.orst.edu/dept/ pnw-erc/

776

S. Gregory

References Baker, J.P., Hulse, D.W., Gregory, S.V., et al., 2004. Alternative futures for the Willamette river basin. Ecol. Appl. 14, 313–324. Gregory, K.J., Gurnell, A.M., 1988. Vegetation and river channel form and process. In: Viles, H.A. (Ed.), Biogeomorphology. Basil Blackwell, Oxford, pp. 11–42. Gregory, K.J., 1992. Vegetation and river channel process interactions. In: Boon, P.J., Calow, P., Petts, G.E. (Eds), River Conservation and Management. John Wiley and Sons, Chichester, UK, pp. 255–269. Gregory, S., Ashkenas, L., Oetter, D., et al., 2002a. Mainstem river. In: Hulse, D.W., Gregory, S.V., and Baker, J.P. (Eds), Willamette River Basin: Trajectories of Environmental and Ecological Change. Oregon State University Press, Corvallis, OR, pp. 112–113. Gregory, S., Ashkenas, L., Oetter, D., et al., 2002b. Longitudinal patterns – channel. In: Hulse, D.W., Gregory, S.V., and Baker, J.P. (Eds), Willamette River Basin: Trajectories of Environmental and Ecological Change. Oregon State University Press, Corvallis, OR, pp. 134–135. Gregory, S., Ashkenas, L., Oetter, D., Wildman, K., 2002c. Revetments. In: Hulse, D.W., Gregory, S.V., and Baker, J.P. (Eds), Willamette River Basin: Trajectories of Environmental and Ecological Change. Oregon State University Press, Corvallis, OR, pp. 138–139. Gregory, S.V., Ashkenas, L.R., Minear, P., 2002d. Application of analysis of historical channel change in the restoration of large rivers. Verhandlungen der Internationale Vereinigung Limnologie 27 (7), 4077–4086. Gregory, S.V., Bisson, P.A., 1996. Degradation and loss of anadromous salmonid habitat in the Paciﬁc Northwest. In: Stouder, D. and Naiman, R.J. (Eds), Paciﬁc Salmon and Their Ecosystems: Status and Future Options. Chapman Hall, pp. 277–314. Gregory, S.V., Hulse, D.W., Landers, D.H., Whitelaw, E., 1998. Integration of biophysical and socioeconomic patterns in riparian restoration of large rivers. In: Wheater, H. and Kirby, C. (Eds), Proceedings of the British Hydrological Society International Conference, Exeter: Hydrology in a Changing Environment. Wiley, Chichester, UK, pp. 231–247. Gurnell, A., Petts, G., 1995. Changing River Channels. Wiley, Chichister, UK, 442pp. Hulse, D., Gregory, S.V., 2001. Alternative futures as an integrative framework for riparian restoration of large rivers. In: Dale, V.H. and Haeuber, R. (Eds), Applying Ecological Principles to Land Management. Springer-Verlag, New York, NY, pp. 194–212. Hulse, D., Gregory, S.V., 2004. Integrating resilience into ﬂoodplain restoration. Urban Ecosyst. 7, 295–314. Hulse, D., Gregory, S.V., Baker, J. (Eds), 2002. Willamette River Basin Planning Atlas: Trajectories of Environmental and Ecological Change. Oregon State University Press, Corvallis, OR, p. 178. Hyman, J.B., Leibowitz, S.G., 2000. A general framework for prioritizing land units for ecological protection and restoration. Environment. Manage. 25 (1), 23–35. Interagency Floodplain Management Review Committee (IFMRC), 1994. Sharing the challenge: ﬂoodplain management into the 21st century. Washington, DC. Klingeman, P.C., 1973. Indications of streambed degradation in the Willamette Valley. Project completion report (OWRR Project A-016-ORE) submitted to Ofﬁce of Water Research, U.S. Department of Interior. 99pp. Ligon, F.K., Dietrich, W.E., Trush, W.J., 1995. Downstream ecological effects of dams. BioScience 45 (3), 183–192. Magilligan, F., Nislow, K.H., Graber, B.E., 2003. Scale-independent assessment of discharge reduction and riparian disconnectivity following ﬂow regulation by dams. Geology 31 (7), 569–572. Naiman, R.J., Decamps, H., 1990. The Ecology and Management of Aquatic–Terrestrial Ecotones. UNESCO-MAB, Paris. Naiman, R.J., Decamps, H., Pastor, J., Johnston, C.A., 1988. The potential importance of boundaries to ﬂuvial ecosystems. J. N. Am. Benthol. Soc. 7, 289–306. National Research Council (NRC), 1992. Restoration of Aquatic Ecosystems. National Academy Press, Washington, DC. Petts, G. and Foster, I., 1985. Channel morphology. Pages 140–174 In: River and Landscape. Edward Arnold Publishers, London, 275pp.

Historical channel modification and floodplain forest decline

777

Petts, G.E., 1990. The role of ecotones in aquatic landscape management. In: Naiman, R.J. and Decamps, H. (Eds), The Ecology and Management of Aquatic–Terrestrial Ecotones. UNESCO, Paris/Parthenon, Carnforth, UK, pp. 227–261. Pie´gay, H., 1998. Interactions between large woody debris and meander cutoff (example of the Mollon Site on the Ain River, France). Z. Geomorphol. 42, 187–208. Pie´gay, H., Bravard, J.P., 1997. Response of a Mediterranean riparian forest to a 1 in 400 year ﬂood, Ouveze River, Drome-Vaucluse, France. Earth Surf. Process. Landf. 22 (1), 31–43. Schumm, S.A., 1968. River adjustment to altered hydrological regime – Murrumbidgee River and paleochannels, Australia. United States Geological Survey, Professional Paper 598. 62pp. Swanson, F.J., Johnson, S.L., Gregory, S.V., Acker, S.A., 1998. Flood disturbance in a forested mountain landscape. BioSci. 48 (9), 681–689. Van Sickle, J., Baker, J., Herlihy, A., et al., 2004. Projecting the biological condition of streams under alternative scenarios of human land use. Ecol. Appl. 14, 368–380.

Gravel-Bed Rivers VI: From Process Understanding to River Restoration H. Habersack, H. Pie´gay, M. Rinaldi, Editors r 2008 Elsevier B.V. All rights reserved.

779

30 Restoring riverine landscapes at the Drau River: successes and deficits in the context of ecological integrity Susanne Muhar, Mathias Jungwirth, Gu¨nther Unfer, Christian Wiesner, Michaela Poppe, Stefan Schmutz, Severin Hohensinner and Helmut Habersack

Abstract In the 19th and 20th centuries, most alluvial rivers in the northern hemisphere were severely disturbed with respect to their hydro-morphology, connectivity, and bedload and discharge regimes. In Austria the relative frequency of braided reaches declined from 28% to 1% over the last century. Thus, several recent restoration efforts have concentrated on formerly dynamic gravel-bed rivers affected by channelisation and river-bed degradation. This paper examines the successes and constraints of selected restoration examples of different spatial extent along the Drau River, Carinthia. Both the hydro-morphological conditions and the status of the ﬁsh fauna are assessed using a 5-tiered scheme according to the EU Water Framework Directive (WFD) based on the type-speciﬁc physical environment. The results show clear improvements of the habitat and ﬁsh ecological situation in rehabilitated sites of the Drau River. In particular, juvenile stages of the key ﬁsh species – the grayling (Thymallus thymallus L.) – beneﬁt from increased areas of shallow habitats; the ecological status improved between 0.2 and 0.9 ecological classes according to the WFD, depending on the spatial extent of the measures. Despite increased efforts in habitat rehabilitation, restoration success is still limited by remaining ecological deficits, such as the disrupted longitudinal river continuum and hydro-peaking, which were not addressed in the project. The presented analyses yield a better perspective on major ecological requirements for future restoration efforts of alluvial riverine landscapes.

E-mail address: [email protected] (S. Muhar) ISSN: 0928-2025

DOI: 10.1016/S0928-2025(07)11164-0

S. Muhar et al.

780 1.

Introduction

In the 19th and especially in the 20th centuries, many alluvial rivers in the northern hemisphere were severely disturbed by anthropogenic impacts related primarily to improved ﬂood protection and navigation as well as to energy production and irrigation. Out of the 139 largest river systems of North America, Europe and the republics of the former Soviet Union, the hydrological regime of 77% has to be classiﬁed as moderately or strongly altered (Dynesius and Nilsson, 1994). Focussing on the alpine region, Martinet and Dubost (1992) reported that only 10% of the river reaches can be classiﬁed as ‘‘near natural’’, an assessment also supported by vegetation studies on alpine braided rivers (Mu¨ller, 1991). In Austria, an investigation of the 53 largest rivers with catchments larger than 500 km2 (in total 5265 river kilometres) showed that about 80% (4200 km) are moderately or heavily impacted by human activity and thus no longer correspond to their original channel morphology and hydrology (Muhar et al., 2000). Prior to systematic channel regulation programmes in the 20th century, the rivers of the alpine region were characterised by a large set of different aquatic and semi-aquatic habitats due to the diverse geomorphologic conditions. Originally, 1500 km of braided reaches represented about 28% of the large river systems in Austria (compare Muhar et al., 2000 and Fig. 30.1) of those only 64 km are still largely undisturbed. These data underline (1) the enormous degree to which originally braided rivers have been altered, (2) the urgent need to protect the last river sections corresponding to this morphological type

morphological character prior to systematic channel regulation current morphological character 1600 28%

28%

1400

24% 23%

river kilometers

1200 19%

18%

1000

15%

800

13%

11%

600

8% 400

6%

5%

200

1%

0 constrained incised meander

braided

pendulous meander

bending

linear

large impoundment

Figure 30.1. Morphological character of the large Austrian rivers (catchment areas 4500 km2) before and after systematic channel regulation (53 rivers, 5265 km length). Out of the originally ﬁve morphological types (hatched bars), a high proportion is currently lost or strongly altered by water engineering measures, representing artiﬁcial morphological characters (e.g., bending, linear).

Restoring riverine landscapes at the Drau River

781

and (3) the need for a comprehensive restoration programme which focus on the restoration of alluvial river landscapes at the catchment scale. In Europe, ﬁrst attempts to improve the physical environment of rivers started in the 1980s with different types of measures to enhance instream structures at the local or reach scale (Jungwirth et al., 1993; Muhar, 1996; Cowx and Welcomme, 1998). More or less concurrently, similar approaches were undertaken in other regions worldwide, as described in a literature review edited by the FAO (Roni et al., 2005). Since then, restoration has gradually progressed from a simple re-structuring of the river bed and the riparian zone to an enhancement of natural attributes that have been measurably degraded (Stanford and Ward, 2001) at the reach or even catchment scale. Roni et al. (2005) point out these fundamental differences in restoration approaches by distinguishing between (1) restoring natural processes and (2) manipulating or enhancing habitats. Though it is not always easy to classify realised projects, most would likely be assigned to this second category. Bradshaw (1997), Stanford et al. (1996), Middleton (1999) and Jungwirth et al. (2002) apply a more differentiated terminology by using the terms enhancement, improvement, mitigation, creation or rehabilitation of habitats. Henry and Amoros (1995) suggest to deﬁne ‘‘restoration as returning an ecosystem to its conditions prior to disturbance (if known and possible), or, as in most cases, to a state as similar as possible to that which prevailed prior to disturbance, according to the changes that have occurred in the watershed’’. In those cases where attempts are limited to partial restoration or to artiﬁcial simulation of natural processes or structures, the term rehabilitation should be used. Even today, most restoration projects continue to focus much more on habitat rehabilitation and on re-establishing longitudinal connectivity. The frequently demanded comprehensive restoration approach, taking a ‘‘long-term and large-scale perspective’’ (Pie´gay et al., 2000), is rarely applied. This typically reﬂects less a lack of knowledge than restrictive legal, administrative or ﬁnancial framework conditions. Only recently more comprehensive measures are being implemented. In this paper we deal with such a river restoration programme at the Drau River in the Austrian Province of Carinthia, which comprised rehabilitation measures of different spatial extent. Beyond initiating type-speciﬁc habitats, the programme aimed to go beyond ‘‘standard’’ restoration approaches: it strived to re-introduce morphodynamic processes and developing and reshaping type-speciﬁc aquatic and semi-terrestric habitats, at least at the local to reach scale. The programme started in the 1990s and is still continuing in 2007. The realised measures were supported by national funds and the EU LIFE-Nature programme and are directed at various goals: the ﬁrst years concentrated on rehabilitation measures in the context of ﬂood protection requirements. By the late 1990s the focus changed to improving the ecological status of the river according to the speciﬁcations of the EU Water Framework Directive (WFD) as well as the EU Bird Protection and Fauna-Flora-Habitat Directives. In contrast to numerous restoration programmes, which lack a systematic monitoring of effectiveness, the different rehabilitated river stretches at the Drau River were repeatedly assessed with regard to habitat conditions, ﬁsh fauna, vegetation, birds, ripicolous arachnid and beetle faunas, etc. (Kucher et al., 2003; O¨koteam, 2003; Unfer et al., 2004b). This helped to evaluate the success or failure of the rehabilitation efforts and also provided knowledge transfer and beneﬁts for new projects (Caruso, 2006).

782

S. Muhar et al.

The methodological approach of the assessment procedure applied in this case study differed considerably from those that deﬁne evaluation as a mere comparison between the degraded, channelised status and the post-rehabilitation status. In addition to analysing these different states, the monitoring data were related to typespeciﬁc reference conditions in the sense of a guiding view and assessed accordingly. The importance of baseline or reference data for evaluating restoration success or failure has been pointed out by numerous authors (Henry and Amoros, 1995; Jungwirth et al., 2002). Determining reference conditions is clearly a crucial issue in river restoration and monitoring and therefore the subject of many scientiﬁc discussions. Even before the WFD deﬁned a normative scheme to evaluate the deviations in ecological quality from a pre-disturbance state, expressed as ‘‘high ecological status’’, the question of how to establish reference conditions was a key issue in river restoration and monitoring programmes, particularly in Central Europe (Kern, 1992; Muhar et al., 1995, 2000). At that time the ‘‘Leitbild’’-concept as a guiding image was developed. It referred to the type-speciﬁc character of an undisturbed or nearly undisturbed river system (Hughes, 1995). Due to the loss of pristine river sites, methods to quantitatively outline such reference conditions are often restricted to the analysis of historical data and the development of models. Additional expert judgement may be necessary to interpret or complement these data. Accordingly, reference conditions can be successfully reconstructed when historical maps, land surveys and hydrographs (compare also Galat and Lipkin, 2000; Hohensinner et al., 2005a), ﬁsheries reports, etc. are available. Such data assist to describe the general type of the river-ﬂoodplain system, to deﬁne the extension of the active channel and the ﬂoodplains and to analyse the morphological character of the river landscape and its habitat attributes, including in many cases also the ﬁsh fauna as a biological key component (Haidvogl and Waidbacher, 1997; Schmutz et al., 2000). When using historic information for deﬁning reference conditions it has to be regarded, that major boundary conditions like sediment regime, land use or rainfall have changed over long time periods. Thus, processes like morphodynamics, derived from a sequence of historic maps, have to be described and analysed within the boundary conditions, given at that time. Henry and Amoros (1995) also underline that river restoration projects rely upon the knowledge of pre-disturbance conditions and refer again to methods involving old aerial photographs and maps, historical records, or soil core samples. At the same time, these authors note the impossibility of reproducing the pre-perturbation state based on hydrology, water quality or sediment regime. Concerning the Austrian rivers, these data often originate from periods between 1800 and 1850 and generally correspond to the situation prior to systematic river engineering measures or the construction of large hydroelectric stations. They refer back to a notion of ‘‘organically developed’’ Central European cultural landscapes in which the fundamental functions and system-inherent processes of most river systems still were largely intact. This general methodological approach of the Leitbild concept has become widely accepted (Kern, 1992; Hughes, 1995; Hughes et al., 2000). Nijboer et al. (2004) also support this concept and state that for the ‘‘purposes of the WFD, ‘undisturbed conditions’ may be interpreted as being those existing before the onset of intensive agriculture or forestry and before large-scale industrial

Restoring riverine landscapes at the Drau River

783

disturbances. In many areas in northern Europe this would correspond to a time period around the mid-1800s’’ (see also Petts, 1989). For this article we selected ﬁve rehabilitated river stretches from the whole project area. They represent different spatial dimensions of restoration approaches and document the assessment methods and results regarding hydro-morphology and ﬁsh ecology. Currently many EU member states are challenged to implement the new legal regulations of the WFD. With this case study as a representative example for grave-bed river rehabilitation in the alpine region, we intend to provide and discuss basic information for further restoration projects at gravel-bed rivers. Our results should help to determine whether merely rehabilitating this increasingly degraded river type succeeded in fulﬁlling the WFD goals of attaining good ecological status. At the same time, they should help to analyse ongoing deficits, which are rarely described in detail in restoration literature.

2.

A river restoration programme: case study of the Drau River

The restoration programme at the Upper Drau River, Carinthia, Austria, will be used to document the successes as well as the constraints and ongoing deficits of various rehabilitation measures. These results will then be discussed in the context of the goals and requirements of a comprehensive restoration approach. 2.1.

General description of the study area

The Upper Drau Valley forms the geological frontier between the Crystalline of the Central Alps and the Southern Limestone Alps near the border to Italy. The investigation area comprises a 70-km-long section of the Drau River, a 6th order Alpine river with a nivo-glacial hydrological regime, a mean annual discharge of approximately 70 m3 s1 and a 100-yr ﬂood event of nearly 1000 m3 s1. The ﬁrst major river stabilisation measures on the Drau River were undertaken in the mid-19th century. From 1930 onwards the river was continuously regulated. Typical river elements such as side arms, gravel bars and ﬂoodplains were widely reduced. At the same time the adjacent ﬂoodplains were successively used for intensive agriculture and forestry. Settlements and infrastructure began to spread over the former ﬂood retention area. Finally, the originally pendulous and in sections braiding river became a uniformly broad, monotonous running water. The narrow channel, together with bedload retention in the upper catchment caused by torrent control structures and impoundments, lack of side erosion, increase of bed slope and significant gravel mining until 1993, led to the well-known phenomenon of river-bed degradation (average 0.6 m in 60 yr, maximum 1.5 m, Habersack and Nachtnebel, 1995). As a consequence the levels of the active channel and its former ﬂoodplains were decoupled. Lateral connectivity decreased dramatically and resulted in largely altered habitat conditions for both the aquatic and the ﬂoodplain biota. The river course downstream of the study area is characterised by a chain of hydropower plants. Since around 1980 the lowermost part of the study area is

S. Muhar et al.

784

hydro-peaking

bedload trapping in the upper catchment Kleblach I

Spittal

hydro-peaking

Dellach Greifenburg

Kleblach II

flood protection & channel regulation 10 km

Figure 30.2. Overview of the Upper Drau River Valley: localisation of human impacts (channel regulation, hydro-peaking, bedload trapping) and rehabilitated river sections.

additionally inﬂuenced by hydro-peaking of a reservoir power station. The surges (up to 85 m3 s1) lead to water-level ﬂuctuations of more than 70 cm during low ﬂow. In particular, during the autumn/winter season with a base ﬂow of 25–30 m3 s1, the riparian areas and the water/land transition zones are severely affected by frequent water level ﬂuctuations (Fig. 30.2). Despite the severe human alterations and uses, the river and its ﬂoodplains still exhibit a high regeneration potential. Various alder and ash communities are present along the river featuring the largest remaining inner alpine ﬂoodplain forests of this vegetation type in Austria. They have therefore been classiﬁed as ‘‘priority habitats’’ according to the European Fauna-Flora-Habital-Directive (Council Directive 92/43/ EEC). Remnants of the former channel network, wetlands and different types of ﬂoodplain habitats within the active zone which still feature morphodynamic processes harbour rare and endangered plant and animal species, e.g., the tamarisk (Myricaria germanica L.), the Ukrainian brook lamprey (Eudontomyzon mariae Berg) and the Danube salmon (Hucho hucho L.).

2.2.

Restoration programme

First restructuring measures at the Dray River began in the early 1990s and involved removing the riprap along the shoreline, widening the river bed and initiating typespeciﬁc instream structures (Habersack and Nachtnebel, 1995; Michor et al., 1998).

Restoring riverine landscapes at the Drau River

785

Due to the successful habitat improvement and the increasing public acceptance of these ﬁrst rehabilitation efforts, more comprehensive measures followed. They increasingly allowed morphodynamic processes such as lateral erosion and sedimentation, which created and developed pioneer sites and type-speciﬁc habitats in general (Jungwirth et al., 2002). An overriding goal was to stop river-bed degradation and to reconnect the decoupled river and ﬂoodplain levels. Accordingly, the most important measure was river-bed widening to reduce bed shear stress as well as to increase lateral erosion processes and sedimentation of bedload. This, in turn, led to an aggradation of the river bed and to improved hydrological connectivity (Habersack and Pie´gay, this volume). The objective of the subsequent LIFE–Nature project ‘‘Restoration of the wetland and riparian area at the Upper Drau River’’ (1999–2003) was large-scale restoration of the river and its ﬂoodplain forests with a clear focus on improving river morphology and connectivity. Mitigation measures regarding the disturbed hydrological regime (see the result part) were not addressed. In the following we deal with the different types and dimensions of rehabilitation measures implemented between 1991 and 2003. Five different sites were selected (compare Figs. 30.2 and 30.3). The ﬁrst set of rehabilitation sites included Kleblach I (total length 400 m, river-bed widening on one bank) and Dellach (250 m, river-bed widening on one bank). The second series included Greifenburg (total length 1000 m, river-bed widening on both banks), Kleblach II (total length 1900 m, river-bed widening on both banks plus reconstruction of a large former side channel) and Spittal (total length 2000 m, river-bed widening on both banks). These rehabilitation measures at the local to reach scale initiated dynamic processes in terms of lateral erosion, sedimentation, sediment transport, channel braiding as well as ﬂooding of the riverbanks and the adjacent areas. As a consequence, type-speciﬁc aquatic and semi-aquatic habits – e.g., side-arms, rifﬂe areas, sediment bars and islands, steep erosion banks as well as low gradient riparian zones and water bodies with different degree of connectivity to the main channel – developed, representing those physical components which originally characterised the unimpaired riverine landscape (compare Section 2.4.1). Concerning the scale relevance (Habersack, 2000), all these measures depend on the overall catchment-wide boundary conditions. Thereby, the existing bedload input is a prerequisite for the functioning of river-bed widening as a measure against bed degradation (Habersack and Pie`gay, this volume). Furthermore, within the reach scale the length and width dimensions of the widening measure determine its success. Numerical modelling and ﬁelds studies showed that a length of the widening sections less than about 600 m does not lead to the intended morphodynamics at the Drau River (Habersack et al., 2000, 2003). This means that the Dellach site, whose length was purposefully limited to avoid excessive aggradation because of ﬂood risk, Kleblach I and even Greifenburg are true local to point scale measures; Kleblach II and Spittal, however, are local to reach scale measures, which are able to initiate self-forming morphodynamics. At the River Drau the connection to the bedload sources in torrential catchments is widely given, so that according to sediment transport measurements (Habersack and Laronne, 2002) still enough bedload input to the restored reaches with river-bed widening exists. Nevertheless, a partial reduction of total bedload input is caused by torrent control structures and weirs of hydropower plants.

S. Muhar et al.

786

(a) channelisedDrau River

(b) rehabilitated site “Dellach”

(c) rehabilitated site “Kleblach I”

(d) rehabilitated site “Greifenburg”

(e) rehabilitated site “Kleblach II”

(f) rehabilitated site “Spittal”, still influenced by hydro-peaking)

Figure 30.3. The channelised Drau River (a) compared to ﬁve sites with different rehabilitation measures: one-sided, small-scale river-bed widening at Dellach (b), newly created side arm at Kleblach I (c), bothsided river-bed widening at Greifenburg, initiating new sediment bars (d), large-scale river-bed widening allowing lateral dynamics and the development of different water bodies (large side arms, backwaters) at Kleblach II (e), destabilised and ﬂattened water – land transition areas as Spittal (f).

The effects of the rehabilitation measures were assessed within a 5-year monitoring programme. Beside the ﬁve rehabilitated river stretches documented in this paper, additional measures were carried out over the last few years and restoration at a larger spatial scale will be continued within the framework of a new EU-LIFE-Nature project ‘‘Life in Upper Drau River’’ which started in late 2006.

Restoring riverine landscapes at the Drau River 2.3. 2.3.1.

787

Assessment methods General methodological aspects

As a general methodological outline, WFD speciﬁes a normative scheme, which deﬁnes ﬁve classes of the ecological status: Setting the ‘‘high ecological status’’ as the reference level, categories 2 to 5 (‘‘good, ‘‘moderate’’, ‘‘poor’’ and ‘‘bad status’’) document the gradual deviation from this reference level (Directive 2000/60/EEC). This scheme was used to evaluate the effects of rehabilitation at the different sites. For the monitoring design various indicator groups were deﬁned. This contribution focuses on two selected aspects: (1) hydro-morphological parameters help describe the characteristics of the physical environment; (2) ﬁsh are used as key indicators for the aquatic ecosystem. In order to develop appropriate restoration measures as well as to evaluate the state of the riverine landscape before and after restoration, reference data are required. Various methodological approaches are available to describe river-typespeciﬁc habitat conditions and the related coenoses in terms of ‘‘high ecological states’’ (according to the WFD). Many of these approaches are complementary (Palmer et al., 2005). The present Drau River project uses the Leitbild concept as described by Kern (1992), Hughes (1995), Muhar (1996) and Jungwirth et al. (2002). To deﬁne the habitat conditions and ﬁsh communities prior to systematic channel regulation or disturbances of the hydrological regime, all available information on the river and its ﬂoodplain was compiled. This included historical maps and historical ﬁsheries reports, as well as data from reference sites (ﬁeld data from river sites of the same river type) and reference models. Base data for the description of the hydro-morphological reference conditions of the Upper Drau River were a land survey map (Fraziszeische Landesaufnahme) from 1832 to 1834, a survey map from a railway construction project in 1900, a map of the Drau River (‘‘Drauﬂusskarte’’) from 1912 and a historical description of the Drau River by Schmidt (1880). First, the river-type-speciﬁc habitats of the reference situation of the Upper Drau River were qualitatively characterised by verbal description of the ‘‘key habitat types’’. To provide quantitative data, the areal proportion of the three functional zones (low ﬂow channel, gravel bars and ﬂoodplain area) was calculated. Based on the historical data as well as on ﬁeld data from comparable natural reference sites of a gravel-bed river (Gail) south of the Drau valley, and on expert judgement, the range of expected values (minus the maximum values) for the proportion of each habitat type within these three zones was deﬁned. To deﬁne the natural ﬁsh fauna of the Drau River we analysed various historical sources (e.g., Heller, 1871; N.N., 1883; Hartmann, 1898). We also interviewed contemporary witnesses who had already ﬁshed or had a close relationship to the ﬁshes of the Drau River prior to systematic river regulation in the mid-20th century. The oldest of those contemporary witnesses was born in 1902 (Unfer et al., 2004a). The interviews were used carefully to more precisely render the species frequency classes (see Table 30.1), especially for barbell and nase.

S. Muhar et al.

788

Table 30.1. Historically documented ﬁsh species of the upper Drau River prior to systematic channelisation and impacted hydrology. Frequency class Fish species Eudontomyzon mariae (Berg, 1931) Esox lucius (Linnaeus, 1758) Thymallus thymallus (Linnaeus, 1758) Hucho hucho (Linnaeus, 1758) Salmo trutta forma fario (Linnaeus, 1758) Abrams brama (Linnaeus, 1758) Albumus albumus (Linnaeus, 1758) Barbus barbus (Linnaeus, 1758) Carassius carassius (Linnaeus, 1758) Chondrostoma nasus (Linnaeus, 1758) Gobio gobio (Linnaeus, 1758) Leuciscus cephalus (Linnaeus, 1758) Leuciscus souffia agassizi (Valenciennes, in Cuvier and Valenciennes, 1844) Phoxinus phoxinus (Linnaeus, 1758) Rutilus rutilus (Linnaeus, 1758) Scardinius erythrophthalmus (Linnaeus, 1758) Tinca tinca (Linnaeus, 1758) Barbatula barbatula (Linnaeus, 1758) Cobitis taenia (Linnaeus, 1758) Lota lota (Linnaeus, 1758) Cottus gobio (Linnaeus, 1758) Perca fluviatilis (Linnaeus, 1758)

Single

Frequent

Predominant

Ukrainian brook lamprey (Northern) Pike (European) Grayling Huchen, Danube salmon Brown trout Common bream Bleak Barbel Crucian carp Nase Gudgeon (European) Chub Souﬁe

(Eurasian) Minnow Roach Rudd Tench Stone loach Spined loach Burbot Bullhead, sculpin (Eurasian) Perch

Note: The species are classiﬁed according to three frequency classes (single/frequent/predominant).

The assessment procedure involved comparing (1) the ecological status of the degraded, channelised status with the post-rehabilitation states and (2) each of these different states to type-speciﬁc reference conditions. This was done by numerically evaluating each criterion using scores between 1 and 5 (see above). Then, the value of

Restoring riverine landscapes at the Drau River

789

the ‘‘hydro-morphological status’’ and the ‘‘ﬁsh ecological status’’ was deﬁned as the mean value of the scores for each criterion. Finally, the overall ecological status was calculated as the mean value of both assessment groups.

2.3.2.

Methods: habitat assessment

The assessment of the habitat conditions was based on an area-wide habitat survey of the ﬁve rehabilitated river sites and the channelised section (differentiated into low ﬂow channel, gravel bars and ﬂoodplain area). It also incorporated detailed investigations of representative transects within those sites. Habitat conditions were speciﬁed using a list of numerous habitat types and habitat attributes (pools, rifﬂes, stagnant water areas, sediment bars, vegetation structures, etc.) within the active channel, comprising the low ﬂow channel and the aquatic terrestric transition zone up to the level of the 1-year ﬂood event. The ﬁeld survey was carried out during low ﬂow conditions of the river in winter/spring 1998/1999 and 2002/2003 by ﬁeld inspection from the banks. The basis for the ﬁeld mapping was a construction plan of the restoration project (scale 1:500) as well as aerial photos (scale 1:1000). For transect mapping a specifically designed ﬁeld sheet was used. The assessment of the hydro-morphological conditions followed a 5-tiered evaluation scheme (Jungwirth et al., 2002). This was developed in the context of the monitoring programme according to the normative demands of the WFD (see above). For the assessment process nine criteria were used: (1) morphological river type, (2) ratio of the area of the low ﬂow channel, gravel bars and ﬂoodplain area, (3) number of habitat types, (4) extension of habitat types, (5) substrate pattern, (6) hydrological regime, (7) ﬂood dynamics, (8) morphodynamics and (9) groundwater inﬂuence.

2.3.3.

Methods: fish ecological assessment

Data of young-of-the-year grayling (0+ ﬁsh/ﬁrst year class) were collected by semiquantitative electro-ﬁshing (CPUE) in different types of mesohabitats (see Section 2.3.2) in summer 2003. In total, 32 riprap and gravel habitats were sampled in channelised and rehabilitated sections, representing 3245 m of sampled shoreline. The survey was restricted to the measures carried out in the frame of the LIFE Nature project Kleblach II, Dellach and Spittal. Greifenburg and Kleblach I were not sampled in this survey. Quantitative data of the ﬁsh stock of the Upper Drau River were collected applying the strip-ﬁshing method (Schmutz et al., 2001) in autumn 2002. Fish stocks are quantiﬁed by stratiﬁed sampling of distinct, habitat-speciﬁc ‘‘strips’’ with electro-ﬁshing boats, i.e., (1) glides next to the bank, (2) glides shifted from the bank, (3) mid-river habitats, (4) undercut banks and (5) shifted undercut banks. In total we sampled 12,520 m in channelised and 8442 m in restored sections. All mentioned rehabilitation sites were included in the sampling survey, which was done in late October 2002. Single strips were assigned to either a rehabilitated site or the channelised river section. This yielded separate length frequency plots according to this graduation.

S. Muhar et al.

790

To assess the ﬁsh ecological status of different sections we used the multi-level concept for ﬁsh-based, river-type-speciﬁc assessment of ecological integrity published by Schmutz et al. (2000). For the assessment seven criteria are used: (1) rivertype-speciﬁc species, (2) species with self-sustaining populations, (3) ﬁsh region, (4) number of guilds, (5) guild composition, (6) population size and (7) population age structure. Deviations from the undisturbed reference condition are assessed. The ﬁnal assessment procedure involved comparing the actual situation with the reference conditions using a 5-tiered normative scheme (see Section 2.3.1 and Schmutz et al., 2000). 2.4. 2.4.1.

Results Description of the reference conditions (‘‘Leitbild’’)

The physical environment of the natural Drau River can be characterised as follows: the valley bottom has a width up to 500 m and provides room for the development of a pendulous river course, creating slip off as well as undercut banks. Due to the strong bends of the main channel, deep pools and runs with depths of 3–4 m characterise the aquatic area. In sections where valley morphology limits the discharge area, the Drau River is typically represented by one main river channel with an average width between 50 and 80 m, providing space for the development of single islands. Due to high bedload input of numerous tributaries, other sections of the riverine landscape are braided and exhibit an active channel width up to 200 m. These river sections are characterised by high ﬂuvial dynamics that periodically re-shape or create new aquatic and semi-aquatic habitats. High ﬂood events lead to lateral erosion processes and bedload input. The substrate composition is dominated by gravel (mesolithal, with diameters of 6.3–20 cm; O¨NORM M6232, 1997). Additionally, micro- and macrolithal occur (2–6.3, 20–40 cm) along with local patches of ﬁne substrate (sand and silt). The geomorphological conditions promote large sediment bars and islands as key elements of the riverine habitat pattern. Many are non-vegetated or covered with typical pioneer communities (see below). The sequence of pendulous and braiding river sections and the ‘‘functional context’’ of these morphological units provide the diversity of habitats that the riverine fauna and ﬂora need to meet the different habitat demands during their life cycle. Depending on the speciﬁc morphological shape of the valley proﬁle along the river course, the ﬂoodplains have a maximum lateral dimension of 300–500 m. Typical waterbodies within the ﬂoodplain area include periodically connected side arms, ponds and pools, which have gone through difference successional stages due to biogenic and abiotic siltation. The surface area of these habitats is typically small. A differentiation of the alluvial river landscape into low ﬂow channel, gravel bars and ﬂoodplain area provides an important basis for evaluating the distribution of the main functional components of the ecosystem (see Fig. 30.4). The proportion of these three main components along the riverine landscape has been calculated based on digitalisation and investigation of historical maps and aerial photos as well as on ﬁeld mapping (compare also Section 2.3.1). The ﬁsh ecological Leitbild characterises the Upper Drau River as the Grayling zone (hyporhithral) with the grayling (Thymallus thymallus L.) as key species, accompanied

Restoring riverine landscapes at the Drau River

791

Figure 30.4. Reference conditions of a typical braiding river section of the Upper Drau River, deﬁned by the relation of the low ﬂow channel, emerged gravel bars and ﬂoodplain forests (derived from accurate land survey and construction maps 1832–1912).

by brown trout (Salmo trutta forma fario L.), Danube salmon (H. hucho L.) and souﬁe (Leuciscus souffia agassizi Val.). In the past, further important rheophilic cyprinids (e.g., nase; Chondrostoma nasus L.), stagnophilic species such as rudd (Scardinius erythrophthalmus L.) and crucian carp (Carassius carassius L.) as well as various eurytopic species, inhabiting the different ﬂoodplain habitats, contributed to the relatively high species richness of the overall system (see Table 30.1). In total, the original species composition comprised 22 ﬁsh species from eight ecological guilds. The main purpose for interviewing contemporary witnesses was to clarify the relative abundance of barbel (Batbus barbus) and nase (C. nasus). Only one contemporary witness had seen single barbels in the Upper Drau River, whereas all of them conﬁrmed that the nase was among the most abundant species until the 1980s. We therefore classiﬁed the nase as predominant, while the occurrence of barbel must have been restricted to single individuals. Data on ﬁsh biomasses and densities for the undisturbed historical situation are not available, but a dramatic decrease was observed since the ﬁrst quanti tative investigations in the late 1980s. While the average biomass was 150 kg/ha in 1989, it declined to less than 30 kg/ha in 2002 (Unfer et al., 2004a). We assume that the observed ongoing decrease of the ﬁsh stocks is the consequence of long term effects resulting from river regulations works of past decades, the ongoing river-bed degradation and decoupling of back- and sidewaters as well as biotic factors, e.g., the predation by cormorants. 2.4.2.

Assessment results

The monitoring results at all rehabilitated sites clearly prove an increased variety of type-speciﬁc habitats such as stagnant shallow zones, rifﬂe areas, gravel and sand bars, and pioneer sites. As an example, the aquatic zone in Greifenburg increased by 12% (from 76,000 to 85,800 m2; Fig. 30.5). Compared to pre-rehabilitation, four additional aquatic habitat types were created. The most important are shallow-water areas with both ‘‘rifﬂes’’ and ‘‘stagnant areas’’ as essential habitats for gravel

S. Muhar et al.

792 deep water / fast flow deep water / medium flow

aquatic habitat types

deep water / low flow medium water / fast flow medium water / medium flow medium water / low flow shallow water / fast flow shallow water / medium flow channelised shallow water / low flow

rehabilitated reference (min-max)

stagnant water 0

5000

10000

15000

20000

aquatic area

25000

30000

35000

40000

(m2)

Figure 30.5. Distribution of aquatic habitat types in channelised and rehabilitated sites as well as related to the reference conditions (range between min–max indicated by grey arrows) at Greifenburg. Total aquatic area of the reference site: 114,400 m2; area before rehabilitation: 76,000 m2; area after rehabilitation: 85,800 m2.

spawners and for fry and young-of-the-year ﬁshes (Sempeski and Gaudin, 1995a, b). Moreover, the re-development of these habitat types corresponds to the habitat characteristic of the Leitbild (compare the min–max range indicated by grey arrows, derived from comparable near-natural reference sites), even if the proportional area of the rehabilitated habitat types does not yet fully reach these reference values. An overview of the morphological improvements due to rehabilitation can be derived by calculating the relative area of low ﬂow channel, gravel bars and ﬂoodplain zone (see also Section 3); these proportions are a key attribute for the overall habitat characterisation of braiding river ecosystems. The column-sequence on the right-hand side in Fig. 30.6 shows the typical distribution of the three zones under reference conditions: 16% of the surface area belongs to the water and 12% to the riparian zone. River bed widening and the creation of various new waterbodies (e.g., side arms, back waters) distinctly increased aquatic habitats and re-activated erosion and sedimentation processes to more than a double of the riparian zone compared to the channelised river. This doubling of the aquatic-terrestrial transition area, however, still represents a high deficit compared with the natural extent. These examples (Figs. 30.5 and 30.6) demonstrate the two-perspective evaluation approach of this monitoring, which assesses the improvement relative to (1) the channelised situation and (2) to type-speciﬁc reference conditions (see Section 2.3.2; Palmer et al., 2005). Various results of the ﬁsh ecological surveys show habitat improvements for the natural ﬁsh fauna of the Drau River. Particularly juveniles of the key species, the grayling (T. thymallus L.), beneﬁt from the river-bed rehabilitation. Fig. 30.7 shows the densities of young-of-the-year grayling (0+ ﬁsh) in rehabilitates sections

Restoring riverine landscapes at the Drau River

793

100 90%

90

low flow channel

gravel bars

floodplain area

82% 80 72% 70

area (%)

60 50 40 30 20

16%

14% 8%

10

12% 4%

2% 0 channelised

rehabilitated

reference cond.

Figure 30.6. Proportions of low ﬂow channel, gravel bars and ﬂoodplain areas* of the natural reference (corresponding to high ecological status), compared to the situations in channelised sites and at Kleblach II after rehabilitation (total area: 99.4 ha; compare Fig. 30.3). *Note: Floodplain area in the ‘‘channelised’’ and ‘‘rehabilitated’’ situation is characterised by different vegetation and especially land use types.

120

individuals [CPUE]

100

80

60

40

20

0 channelised Drau River

Dellach

Kleblach II

Spittal

Figure 30.7. Densities of 0+ grayling in channelised sections of the Drau River and within the rehabilitation sites Dellach, Kleblach II and Spittal.

S. Muhar et al.

794

(Dellach, Kleblach II and Spittal) and in channelised stretches of the Drau River. These results derive from the semi-quantitative electro-ﬁshing survey in summer 2003. While the densities in channelised sections, where the banks are paved with riprap, are extremely low (6.1 CPUE), high densities of 0+ grayling were recorded within the rehabilitated sites (39.8–117 CPUE), where the dominant structure type at the banks is shallow gravel bars. Although the 0+ densities at Spittal (39.8 CPUE), which is impacted by hydro-peaking surges, are clearly higher than at the channelised Drau River, the upstream values near Dellach (88.3 CPUE) and Kleblach are distinctly higher. The quantitative survey in the autumn situation in 2002 also highlights the success of the rehabilitation efforts. Biomasses as well as densities of the grayling clearly increased in the rehabilitated versus channelised sites (Fig. 30.8). These increased densities show that especially habitat for early life-stages of the grayling could be initiated (compare Fig. 30.9). In contrast to the grayling, both brown and rainbow trout show higher values in channelised sections. The riprap on the banks favours these species: it provides suitable cover for trout, while the grayling is dependent on the availability of shallow gravel banks (Fig. 30.8). Fig. 30.9 shows the length–frequency distributions of the grayling from the autumn survey in 2002 in regulated and rehabilitated river stretches (with and without inﬂuences of hydropeaking). Due to the extensive channel widening, shallow water areas evolved; they represent the crucial habitats for the juvenile stages of this rheophilic species. A clear peak of juvenile ﬁshes in the length classes 70–160 mm (ﬁrst year class) is evident in rehabilitated sections (Fig. 30.9b) with an undisturbed hydrological regime. In channelised stretches (Fig. 30.9a) this age class is markedly reduced, as is the total number of caught individuals. The lowest values in the autumn situation occur within the restructured section near Spittal (Fig. 30.9c), which is heavily inﬂuenced by hydropeaking. Here, the total number of ﬁsh is reduced to a low level of 115 individuals and the ﬁrst year class is more or less missing. This indicates that hydro-peaking causes inadequate habitat conditions for the 0+ grayling when the water level ﬂuctuations increase in combination with reduced base ﬂows in autumn. As a consequence the actual population structure in the hydro-peaking impacted section is not natural. 2.4.3.

Final evaluation of ecological status

The ﬁnal evaluation represents the average total of all assessment criteria mentioned in Sections 2.3.2 and 2.3.3. The results show a stepwise improvement in the overall ecological status. This improvement primarily reﬂects the spatial dimension of the particular rehabilitation site and the magnitude of re-established morphodynamic processes. Dellach, as one of the small-scale sites, shows an improvement of only 0.2 ecological classes for each indicator group. Kleblach II, achieving a value of 2.1 (habitat) respectively 2.7 (ﬁsh) improved by about a whole quality class (see Table 30.2). This can largely be explained by the threefold enlargement of the active channel, including the development of a large side arm and an island with alder forest. The assessment regarding the ﬁsh ecological improvements proves a positive trend in all the relevant measures. The best results were achieved in the most extensive measure Kleblach II, where in particular the habitat availability for the key species, the grayling, remarkably improved. The same amount of suitable habitat structures were

Restoring riverine landscapes at the Drau River

795

25 channelised Drau River rehabilitated sites

biomass [kg/ha]

20

15

10

5

0 grayling

brown trout

rainbow trout

300 channelised Drau River rehabilitated sites 250

density [ind./ha]

200

150

100

50

0 grayling

brown trout

rainbow trout

Figure 30.8. Biomass (top) and densities (bottom) of the main ﬁsh species in channelised and rehabilitated sections of the Drau River.

created in Spittal, but due to hydro-peaking the age structure and ﬁsh density remain distorted (see Section 2.4.2), reducing the positive effects of habitat rehabilitation. The results demonstrate the following rehabilitation effects on the riverine environment and its biocoenoses: (1) distinct increase in the variety of the aquatic and semi-aquatic habitats of the waterzone and the gravel bars, particularly regarding the

796

Individuals 90

100

220-229 240-249 260-269 280-289 300-309 320-329 360-369

380-389

380-389

400-409

400-409

400-409

420-429

420-429

420-429

440-449

440-449

440-449

460-469

460-469

460-469

480-489

480-489

480-489

>500

>500

>500

S. Muhar et al.

340-349

360-369

380-389

80

160-169

340-349

360-369

70

320-329

60

140-149

Fish length (mm)

300-309

50

120-129

200-209

280-289

40

80-89 100-109

180-189

260-269

30

60-69

200-209 240-249

20

0 40-49

180-189 220-229

10

(a) channelised (n = 403)

90

100

80

70

160-169

60

140-149

50

120-129

40

100-109

30

340-349

80-89

20

320-329

0

300-309

60-69

10

280-289

40-49

Fish length (mm)

260-269

(b) rehabilitated sites Dellach and Kleblach II (n = 487)

240-249

90

220-229

100

200-209

80

180-189

Individuals 70

160-169

60

140-149

50

120-129

40

100-109

30

80-89

20

0 60-69

10

40-49

Fish length (mm)

(c) rehabilitated site Spittal, still hydrologically impacted by hydro-peaking (n = 115)

Figure 30.9. Length frequency of grayling in (a) channelised, (b) rehabilitated and (c) rehabilitated, but still hydrologically impacted sections (due to hydro-peaking) of the Drau River.

Individuals

Restoring riverine landscapes at the Drau River

797

Table 30.2. Ecological evaluation of the channelised river stretch and the rehabilitated sites regarding habitat and ﬁsh.

Habitat Fish

River Drau channelised

Dellach rehabilitated

Kleblach I rehabilitated

Greifenburg rehabilitated

Kleblach II rehabilitated

Spittal rehabilitated

3.2 3.9

3.0 3.7

3.1 3.4

3.0 3.3

2.1 2.7

2.9 3.0

development of type-speciﬁc habitats such as slow-ﬂowing shallow zones, rifﬂe area, gravel and sand bars, and pioneer sites; (2) increased densities and improved population structure of key ﬁsh species (e.g., grayling) as a crucial prerequisite of an improved population structure. The re-initiation of ﬂuvial dynamics and the development of pioneer sites along the gravel bars were proved by other biological investigations not mentioned in Section 2.3.1. For example, characteristic key plant species (e.g., M. germanica) newly colonised the sediment bars and islands and, in general, the riparian areas with pioneer vegetation communities increased considerably (Kucher et al., 2003). Also, the monitoring on the ripicolous arachnid and beetle faunas showed clear improvements (O¨koteam, 2003; Unfer et al., 2004b).

3.

Conclusions and perspectives

Land drainage, ﬂood control by levees, river regulation by hydro-power plants, various alterations of the hydrological regimes due to water diversion, hydro-peaking etc., and especially measures truncating bedload transport followed by river-bed degradation isolate rivers from their ﬂoodplain and have been the major factors behind physical habitat degradation (Petts, 1996; Jungwirth et al., 2002). These impacts heavily alter the type-speciﬁc natural disturbance regimes that normally maintain the dynamic complexes of ecotones and that are the primary physical factors structuring river ecosystems (Ward and Wiens, 2001). The resulting decrease in hydromorphological dynamics, habitat turnover and hydrological connectivity reduces habitat diversity remarkably. This changes such systems from their original ‘‘shiftingmosaic steady-state type’’ (sensu Bormann and Likens, 1979a, b) to an ecologically truncated ‘‘static-state system’’ (Hohensinner et al., 2005a). At the same time those human impacts also limit future restoration programmes. The assessment of formerly braiding rivers in Austria (Section 1) shows that nearly no such river systems remain intact. The braiding river sections of the Drau River are a part of these altered and impacted river landscapes. These results underline both the urgent need for protecting the last remnants of intact systems and the great demand for comprehensive restoration action (Tockner and Stanford, 2002; Palmer et al., 2005). A major issue and challenge in all such efforts will be to re-establish the balanced erosion/sedimentation processes and hydrological connectivity conditions typical for the given natural systems prior to degradation. This has been exempliﬁed by the alluvial Danube ﬂoodplain system described by Hohensinner et al. (2002) and Hohensinner et al. (2005b).

798

S. Muhar et al.

In the case of the Drau River, the restoration concept – boosted by the implementation of the WFD in 2000 – went beyond many ‘‘standard’’ approaches that primarily focus on ‘‘form and structure’’ to also include ‘‘processes’’ that need to be re-established (compare Muhar and Jungwirth, 1998; Kondolf, 2000; Jungwirth et al., 2002). The ‘‘planning philosophy’’ and conceptual design was based on ‘‘physical processes’’, re-enabling ﬂuvial dynamics and connectivity. This will ultimately promote the essential ‘‘biological processes’’, structures and habitats necessary for the key species and ecological guilds of gravel-bed rivers. In our opinion, this general restoration approach is crucial for ensuring that the habitats and the biocoenoses of the rehabilitated sites develop and correspond at least qualitatively to the type-speciﬁc characteristics of the given riverine system. Nevertheless, the long-term sustainability and success of the restoration efforts along the Drau River will heavily depend on the future possibilities to resolve two major problems in the catchment beyond structural alterations and habitat degradation: bedload transport and hydro-peaking. Although the connection to bedload sources still mostly exists, a continuous future reduction of bedload input to the restoration reaches would again promote the process of river-bed degradation, despite widening the river bed. The result: costly, rehabilitated river stretches would once again become decoupled from the ﬂoodplains. In general, the restoration concept and measures presented in this paper can certainly be applied to gravel-bed rivers of comparable hydraulic, sedimentological and morphological characteristics. The second still ongoing pressure decreasing the overall success of the restoration efforts is hydropeaking, affecting the habitat quality of the Upper Drau River. Its consequences as an overwhelming impact become obvious in the presented results. In contrast to ‘‘Klebach II’’, which is largely undisturbed hydrologically, the hydrological regime at ‘‘Spittal’’ is heavily inﬂuenced by hydro-peaking surges. Despite the comparable general availability of different habitat types at both sites, the population structure of the key ﬁsh species, the grayling, remains qualitatively and quantitatively distorted at ‘‘Spittal’’ (Fig. 30.7; Fig. 30.9c). The ﬁsh ecological monitoring thus clearly documents that the restoration efforts were unable to reduce the detrimental effects of hydro-peaking on the river system. Beyond these two major impacts, further habitat deficits must be taken into account: (1) until now, the extensive functional de-coupling of the river channel network and its surrounding ﬂoodplains allowed only comparatively small adjacent areas to be restored in terms of connectivity conditions corresponding to the former (potential) ﬂoodplains (see also Habersack and Pie´gay, this volume), (2) regarding the whole river system, only comparatively few areas of dynamically originated, typespeciﬁc habitats could re-establish and (3) many original migration pathways – as important links between different habitats in both longitudinal and lateral direction – are still blocked. The latter, migratory aspect reﬂects the disrupted longitudinal continuum by a downstream chain of hydropower plants and the still poor lateral connectivity. This has caused the stocks of the formerly important nase (Table 30.1) to become extinct. This species clearly depended on open migration routes between distinct habitats within the middle reaches and the Upper Drau River. These migration routes urgently need to be re-opened. One way to evaluate restoration success is to correlate the ﬁsh ecological evaluation of restored and channelised sections with the corresponding aquatic areas. For the Upper Drau River this correlation yields a perspective for future restoration

Restoring riverine landscapes at the Drau River

799

1

fish ecological status

R2 = 0.81 2 Spittal 3.0

Kleblach II 2.7

Kleblach I 3.4

3

3.3 Greifenburg 4

3.9 channelised

3.7 Dellach

5 0

10

20 30 40 50 60 70 80 % aquatic habitat area (related to reference conditions)

90

100

Figure 30.10. Fish ecological status correlated with the percentage of aquatic habitat area at different rehabilitation sites (100% ¼ reference situation).

programmes (Fig. 30.10). The linear regression shows that, despite the positive trend, even the highest surface area of aquatic habitat at site Kleblach II (94% of the reference condition value) fails to yield a ﬁsh ecological status better than 2.7 (compare Section 2.4.3). This reﬂects the multitude of further deficits of the riverine system at the given status, but note that the restoration efforts will continue in the frame of new LIFE Nature project and that additional measures will be carried out for ﬂood protection purposes. The current population structure of the grayling in channelised and in rehabilitated sections shows that the stock in the Drau River has not beneﬁted satisfactorily (Fig. 30.9a, b). Although older year classes (4170 mm) do not differ between sections, the ﬁrst year class significantly increased within rehabilitated parts, providing a positive perspective. This contribution shows that the ﬁrst stage of measures (few and small sized) failed to sufﬁciently improve the ecological status (Fig. 30.10). In a second stage, however, where additional and larger measures were realised, at least juvenile graylings did beneﬁt (Figs. 30.7 and 30.9b). We assume that further expanding the ‘‘patchwork of isolated small-scale measures’’ to a network of measures will yield improvements for the entire grayling population. In order to regain the ecological integrity of impaired riverine landscapes (‘‘good ecological status’’ according to WFD), future restoration programmes at the Drau River must address problems associated with (1) the bedload regime, (2) the interrupted longitudinal continuum caused by downstream hydropower plants, (3) the intensively used ﬂoodplain area along with the loss of wetlands as well as the reduced lateral connectivity between river and ﬂoodplains and (4) the human-induced disturbances through hydro-peaking. This calls for higher-level programmes that focus on the catchment or sub-catchment scale. Here, WFD opens up a new perspective because the guideline speciﬁes comprehensive efforts to re-establish ‘‘good ecological

800

S. Muhar et al.

status’’, which also includes remote impacts within the whole catchment (jungwirth et al., 2005). This case study underlines the importance of such a broad spatial approach. It also emphasises the necessity of implementing additional measures along the Drau river landscape. Site-to reach-scale river rehabilitation does have the potential to be successful, for example in increasing the availability of appropriate habitats. This was documented in the case of the Drau River, regarding, e.g., improved habitat conditions within the aquatic area for gravel-spawning ﬁsh species, or the development of pioneer vegetation due to the re-establishment of naturally disturbed and re-shaped sediment bars (as shown by Kucher et al., 2003). Comparable examples are highlighted by Hughes et al. (2001). The authors demonstrate that riparian pioneer vegetation and ﬂoodplain woodlands can regenerate at these scales, but will only be self-sustaining over the long term if the restoration concept includes the re-activation of geomorphological processes. This once again underlines the need for a more challenging approach to river management across the catchment. Finally, comprehensively monitoring the Drau River was instrumental in analysing and documenting the various impacts; it was also crucial in determining which impacts were successfully addressed by the different restoration measures and to what degree. Such often neglected monitoring efforts are essential to avoid repeating mistakes and to develop an understanding of how rivers respond to restoration actions (Kondolf, 1998; Sear et al., 1998). The Drau River project is an example for this ‘‘learning process’’ by scientists as well as river managers. Here, over a period of more than 15 years, a step-by-step restoration concept has been developed; a chronology of the performance documents the gradual enhancement of the restoration measures. In conclusion, we can draw an optimistic perspective for the Drau River – in the event that the currently discussed design for future restoration measures comprises, stresses and solves the remaining problems. We identify these as the disturbed sediment budget, the hydrological deficits (in particular hydro-peaking), a more extensive lateral connectivity and large-scale solutions for restoring migration pathways within the Middle and Upper Drau River system.

Acknowledgements The ﬁndings presented in this paper are based on projects funded by the European Community (LIFE Nature ‘‘Restoration of wetland and riparian area at the Upper Drau River’’ 1999–2003) and the Austrian Ministry of Agriculture, Forestry, Environment and Water Management together with the Regional Governments and the local authorities. The authors thank those colleagues and students who supported these investigations by their ﬁeld work as well as Michael Stachowitsch for professional scientiﬁc English proofreading. We would also like to thank the organising committee of the 6th International Gravel-Bed Rivers Workshop in St. Jakob/Austria for having provided the opportunity to attend and contribute to this workshop.

Restoring riverine landscapes at the Drau River

801

References Bormann, F.H., Likens, G.E., 1979a. Patterns and Process in a Forested Ecosystem. Springer-Verlag, New York, 253pp. Bormann, F.H., Likens, G.E., 1979b. Catastrophic disturbance and the steady state in northern hardwood forests. Am. Sci. 67, 660–669. Bradshaw, A.D., 1997. What do we mean by restoration? In: Urbanska, K.M., Webb, N.R., and Edwards, P.J. (Eds), Restoration Ecology and Sustainable Development. Cambridge University Press, Cambridge, pp. 8–14. Caruso, B.S., 2006. Effectiveness of braided, gravel-bed river restoration in the Upper Waitaki Basin, New Zealand. River Res. Appl. 22, 905–922. Council Directive 92/43/EEC of 21 May 1992 on the conservation of natural habitats and of wild fauna and ﬂora (OJ L 206, 22.7.1992, p.7). 57pp. Cowx, I.G and Welcomme, R.L. (eds) (1998). Rehabilitation of Rivers for Fish. Oxford: Fishing News Books, Blackwell Science, 204pp. Directive 2000/60/EC of the European Parliament and of the Council of 23 October 2000 establishing a framework for Community action in the ﬁeld of water policy. 72pp. Dynesius, M., Nilsson, C., 1994. Fragmentation and ﬂow regulation of the river systems in the northern third of the world. Science 266, 753–762. Galat, D.L., Lipkin, R., 2000. Restoring ecological integrity of great rivers: historical hydrographs aid in deﬁning reference conditions for the Missouri River. Hydrobiologia 422–423, 29–48. Habersack, H., 2000. The river-scaling concept (RSC): a basis for ecological assessments. In: Jungwirth, M., Muhar, S., and Schmutz, S. (Eds), Assessing the Ecological integrity of Running Waters. Kluwer Academic Publishers, Dordrecht/Boston/London, pp. 49–60. Habersack, H., Nachtnebel, H.P., 1995. Short term effects of local river restoration on morphology, ﬂow ﬁeld, substrate and biota. Regul. Rivers Res. Manage. 10, 291–301. Habersack, H. and Pie´gay, H., this volume. Challenges in river restoration in the Alps and their surrounding areas. Habersack, H., Schober, S., Formann, E., et al., 2003. Flussmorphologisches Monitoring in Rahmen des LIFE-Projektes ‘‘Obere Drau’’. BMLFUW: 20. Flussbautagung LIFE-SYMPOSIUM: Gewa¨sserbetreuung und die Eu-Wasserrahmenrichtlinie – Umsetzung am Beispiel von LIFE Projekten, Sept. 2003, Spittal/Drau, S. 15–45, Wien; ISBN 3-85 174–47-5. Habersack, H.-M., Koch, K., Nachtnebel, H.-P., 2000. Flussaufweitungen in O¨sterreich – entwicklung, stand und ausblick. Oesterr. Wasser- Abfallwirtsch. 52, 143–153. Habersack, H.M., Laronne, J.B., 2002. Evaluation and improvement of bedload discharge formulas based on Helley-Smith sampling in an alpine gravel bed river. J. Hydraul. Eng. 128 (5), 484–499. Haidvogl, G. and Waidbacher, H., 1997. Ehemalige Fischfauna an ausgewa¨hlten o¨sterreichischen Fliebgewa¨ssern. Abt. f. Hydrobiologie, Univ. f. Bodenkultur, Wien, 85pp. Hartmann, V., 1898. Die Fische Ka¨rntens, Separat-Abdruck aus dem XXV. Jahrbuch des naturhistorischen Landesmuseums von Ka¨rnten, Klagenfurt, Ferd. v Kleinmayer Verlag. Heller, C., 1871. Die Fische Tirols und Vorarlbergs, Separt- Abdruck aus der Ferdinandeums-Zeitschrift vom Jahr 1871, Innsbruck, Wagner’sche Universita¨tsbuchdruckerei. Henry, C.P., Amoros, C., 1995. Restoration ecology of riverine wetlands: I. A scientiﬁc base. Environ. Manage. 19, 891–902. Hohensinner, S., Egger, G., Haidvogl, G., et al., 2002. Hydrological connectivity of a Danube riverﬂoodplain system in the Austrian Machland: changes between 1812 and 1991. In: Tre´molie`res, M. (Ed.), Proceedings of the International Conference ‘‘European Floodplains 2002’’ in Strasbourg, France. Hohensinner, S., Haidvogl, G., Jungwirth, M., et al., 2005b. Historical analysis of habitat turnover and age distributions as a reference for restoration of Austrian Danube ﬂoodplains. River Basin Management III, WIT Transactions on Ecology and the Environment. WIT Press, Ashurst, UK, Vol. 83, pp. 489–502.

802

S. Muhar et al.

Hohensinner, S., Jungwirth, M., Muhar, S., Habersack, H., 2005a. Historical analyses: a foundation for developing and evaluating river-type speciﬁc restoration programs. Int. J. River Basin Manage. 3 (2), 87–96 ISSN-1571–5124. Hughes, R.M., 1995. Deﬁning acceptable biological status by comparing with reference conditions. In: Davis, W.S. and Simon, R.F. (Eds), Biological Assessment and Criteria; Tools for Water Resource Planning and Decision Making. Lewis Publishers, Boca Raton, FL, pp. 31–47. Hughes, R.M., Paulsen, S.G., Stoddard, J.L., 2000. EMAP – surface waters: a national, multiassemblage, probability survey of ecological integrity. Hydrobiologia 422–423 (15), 429–443. Hughes, F.M.R., Adams, W.M., Muller, E., et al., 2001. The importance of different scale processes for the restoration of ﬂoodplain woodlands. Regul. Rivers Res. Manage. 17, 325–345. Jungwirth, M., Haidvogl, G., Hohensinner, S., et al., 2005. Leitbild-speciﬁc measures fort he rehabilitation of the heavily modiﬁed Austrian Danube River. Large Rivers Vol. 15 (1–4), Arch Hydrobiol. Supp. 155/1–4, 17–36. Jungwirth, M., Moog, O., Muhar, S., 1993. Effects of river-bed restructuring on ﬁsh and benthos of a 5th-order stream (Melk, Austria). Regul. Rivers 8, 195–204. Jungwirth, M., Muhar, S., Schmutz, S., 2002. Reestablishing and assessing ecological integrity in riverine landscapes. Freshw. Biol. 47 (4), 867–888. Kern, K., 1992. Rehabilitation of streams in South-West Germany. In: Boon, P.J., Calow, P., and Petts, G.E. (Eds), River Conservation and Management. Wiley, Chichester, UK, pp. 321–336. Kondolf, M., 1998. Lessons learned from river restoration projects in California. Aquat. Conserv. Mar. Freshw. Ecosyst. 8, 39–52. Kondolf, M., 2000. Some suggested guidelines for geomorphic aspects of anadromous salmonid habitat restoration proposals. Restor. Ecol. 8–1, 48–56. Kucher, T., Aigner, S., Egger, G., 2003. LIFE-Projekt ‘‘Auenverbund Oberes Drautal’’ – Monitoring Vegetation. Institute fu¨r O¨kologie and Umweltplanung, Klagenfurt, 113pp. Martinet, F., Dubost, M., 1992. Die letzten naturnahen Alpenﬂu¨sse. CIPRA, Kleine Schriften 11, 10–58. Michor, K., Habersack, H., Nachtnebel, H.-P., et al., 1998. MaXnahmenkatalog zur Umsetzung des Leitbildes an der Oberen Drau. Oesterr. Wasser- Abfallwirtsch. 50 (H.1/2), 57–64. Middleton, B., 1999. Wetland Restoration, Flood Pulsing and Disturbance Dynamics. Wiley, New York. Muhar, S., 1996. Habitat improvement of Austrian rivers with regard to different spatial scales. Regul. Rivers Res. Manage. 12, 471–482. Muhar, S., Jungwirth, M., 1998. Habitat integrity of running waters – assessment criteria ant their biological relevance. Hydrobiologia 386, 195–202. Muhar, S., Schmutz, S., Jungwirth, M., 1995. River restoration concepts – goals and perspectives. Hydrobiologia 303, 183–194. Muhar, S., Schwarz, M., Schmutz, S., and Jungwirth, M., 2000. Identiﬁcation or rivers with high and good habitat quality: methodological approach and applications in Austria. In: Jungwirth, M., Muhar, S., and Schmutz, S. (Eds), Assessing the Ecological Integrity of Running Waters, Hydrobiologia 422/423: 343–358. Mu¨ller, N., 1991. Der Lech. Wandel einer WildﬂuXlanschaft. Augsburger O¨kolog. Schriften 2, 174pp. Nijboer, R.C., Johnson, R.K., Verdonschot, P.F.M., et al., 2004. Establishing reference conditions for European streams. Hydrobiologia 516, 93–107. N.N., 1883. ‘‘Die Fischereiverha¨ltnisse im Oberen Drau-, im Mo¨ll- und Gailthale, einschlieXlich jener des WeiXen-, Millsta¨tter- und Ossiacher Sees im Kornlande Ka¨rnten.’’ Mitteilungen des o¨sterreichischen Fischereivereines. pp. 2–14. O¨koteam, 2003. LIFE Projekt Obere Drau. Monitoring Terrestrische Tierwelt. Spinnen, Weberknechte, Skorpione, Laufka¨fer, Kurzﬂu¨gelka¨fer & Libellen. Projektsbericht im Auftrag des Amtes der Ka¨rntner Landesregierung, 152pp. Palmer, M.A., Bernhardt, E.S., Allan, J.D., et al., 2005. Standards for ecologically successful river restoration. J. Appl. Ecol. 42, 208–217. Petts, G.E., 1989. Historical analysis of ﬂuvial hydrosystems. In: Petts, G.E., Mo¨ller, H., and Roux, A.L. (Eds), Historical Change of Large Alluvial Rivers: Western Europe. Wiley, Chichester, pp. 1–18. Petts, G.E., 1996. Sustaining the ecological integrity of large ﬂoodplain rivers. In: Anderson, M.G., Walling, D.E., and Bates, P.D. (Eds), Floodplain Processes. Wiley, Chichester, UK, pp. 535–551.

Restoring riverine landscapes at the Drau River

803

Pie´gay, H., The´venet, A., Kondolf, G.M., Landon, N., 2000. Physical and human factors inﬂuencing potential ﬁsh habitat distribution along a mountain river, France. Geograﬁs. Annal. Ser. A Phys. Geogr. 82 (1), 121–136. Roni, P., Hanson, K., Beechie, T., et al., 2005. Habitat rehabilitation for inland ﬁsheries. Global review of effectiveness and guidance for rehabilitation of freshwater ecosystems. FAO Fisheries Technical Paper. No. 484. Rome, FAO. 116pp. Schmidt, F., 1880. Die Drauregulierung in Ka¨rnten, Seperatdruck aus. ‘‘O¨sterr.-Ungar. Revue,’’ Wien. Schmutz, S., Kaufmann, M., Vogel, B., et al., 2000. A multi-level concept for ﬁsh- based, rivertype-speciﬁc assessment of ecological integrity. In: Jungwirth, M., Muhar, S., and Schmutz, S. (Eds), Assessing the Ecological Integrity of Running Waters, Hydrobiologia 422/423, 279–289. Schmutz, S., Zauner, G., Eberstaller, J., Jungwirth, M., 2001. Die Streifenbeﬁschungsmethode: Eine Methode zur Quantiﬁzierung von Fischbesta¨nden mittelgroXer FlieXgewa¨sser. Oesterr. WasserAbfallwirtsch. 54, 14–27. Sear, D.A., Briggs, A., Brookes, A., 1998. A prelimininary analysis of the morphological adjustment within and downstream of a lowland river subject to river restoration. Aquat. Conserv. Mar. Freshw. Ecosyst. 8, 167–183. Sempeski, P., Gaudin, P., 1995a. Habitat selection by grayling – I. Spawning habitats. J. Fish Biol. 47, 256–265. Sempeski, P., Gaudin, P., 1995b. Habitat selection by grayling – II. Preliminary results on larval and juvenile daytime habitats. J. Fish Biol. 47, 345–349. Stanford, J.A., Ward, J.V., 2001. Revisiting the serial discontinuity concept. Regulated Rivers-Research & Management 17 (4-5), 303–310. Stanford, J.A., Ward, J.V., Liss, W.J., et al., 1996. A general protocol for restoration of regulated rivers. Regul. Rivers Res. Manage. 12, 391–413. Tockner, K., Stanford, J.A., 2002. Riverine ﬂood plains: present state and future trends. Environ. Conserv. 293, 308–330. Unfer, G., Schmutz, S., Wiesner, C., et al., 2004b. The effects of hydro-peaking on the success of river-restoration measures within the LIFE-project ‘‘Auenverbund Obere Drau’’. In: Diego Garcia de Jalon and Pilar Vizcaino Martinez (Eds), Fifth International Symposium on Ecohydraulics, 12.09.200417.09.2004, Madrid; Proceedings of the Fifth International Conference on Ecohydraulics—Aquatic Habitats: Analysis and Restoration, 1, 741–746. Unfer, G., Wiesner, C., and Jungwirth, M., 2004a. Auenverbund Obere Drau—Fischo¨kologisches Monitoring – Endbericht. Studie im Auftrag des Amts der Ka¨rntner Landesregierung—Abt. 18 – Wasserwirtschaft: 94pp. Ward, J.V., Wiens, J.A., 2001. Ecotones of riverine systems: role and typology, spatio- temporal dynamics, and river regulation. Ecohydrol. Hydrobiol. 1, 25–36.

Discussion by M. Roberts Although braided rivers are a central component in the Leitbild approach to river restoration planning for gravel-bed rivers, it can be more apposite in some environments to consider using the properties of ‘wandering gravel-bed rivers’. This channel planform distinguished by multiple channels, armouring of the channel bottom and forested islands, which are often stable over decadal time periods, can be the most geomorphically appropriate planform to use in mountainous environments (Fig. 30.11). The analysis of historical maps from Alpine regions often reveals the presence of forested islands (e.g., Rhoˆne River near Aoste, France) suggesting that the wandering gravel-bed planform would be the most suitable to use for restoration work.

Restoring riverine landscapes at the Drau River

803

Pie´gay, H., The´venet, A., Kondolf, G.M., Landon, N., 2000. Physical and human factors inﬂuencing potential ﬁsh habitat distribution along a mountain river, France. Geograﬁs. Annal. Ser. A Phys. Geogr. 82 (1), 121–136. Roni, P., Hanson, K., Beechie, T., et al., 2005. Habitat rehabilitation for inland ﬁsheries. Global review of effectiveness and guidance for rehabilitation of freshwater ecosystems. FAO Fisheries Technical Paper. No. 484. Rome, FAO. 116pp. Schmidt, F., 1880. Die Drauregulierung in Ka¨rnten, Seperatdruck aus. ‘‘O¨sterr.-Ungar. Revue,’’ Wien. Schmutz, S., Kaufmann, M., Vogel, B., et al., 2000. A multi-level concept for ﬁsh- based, rivertype-speciﬁc assessment of ecological integrity. In: Jungwirth, M., Muhar, S., and Schmutz, S. (Eds), Assessing the Ecological Integrity of Running Waters, Hydrobiologia 422/423, 279–289. Schmutz, S., Zauner, G., Eberstaller, J., Jungwirth, M., 2001. Die Streifenbeﬁschungsmethode: Eine Methode zur Quantiﬁzierung von Fischbesta¨nden mittelgroXer FlieXgewa¨sser. Oesterr. WasserAbfallwirtsch. 54, 14–27. Sear, D.A., Briggs, A., Brookes, A., 1998. A prelimininary analysis of the morphological adjustment within and downstream of a lowland river subject to river restoration. Aquat. Conserv. Mar. Freshw. Ecosyst. 8, 167–183. Sempeski, P., Gaudin, P., 1995a. Habitat selection by grayling – I. Spawning habitats. J. Fish Biol. 47, 256–265. Sempeski, P., Gaudin, P., 1995b. Habitat selection by grayling – II. Preliminary results on larval and juvenile daytime habitats. J. Fish Biol. 47, 345–349. Stanford, J.A., Ward, J.V., 2001. Revisiting the serial discontinuity concept. Regulated Rivers-Research & Management 17 (4-5), 303–310. Stanford, J.A., Ward, J.V., Liss, W.J., et al., 1996. A general protocol for restoration of regulated rivers. Regul. Rivers Res. Manage. 12, 391–413. Tockner, K., Stanford, J.A., 2002. Riverine ﬂood plains: present state and future trends. Environ. Conserv. 293, 308–330. Unfer, G., Schmutz, S., Wiesner, C., et al., 2004b. The effects of hydro-peaking on the success of river-restoration measures within the LIFE-project ‘‘Auenverbund Obere Drau’’. In: Diego Garcia de Jalon and Pilar Vizcaino Martinez (Eds), Fifth International Symposium on Ecohydraulics, 12.09.200417.09.2004, Madrid; Proceedings of the Fifth International Conference on Ecohydraulics—Aquatic Habitats: Analysis and Restoration, 1, 741–746. Unfer, G., Wiesner, C., and Jungwirth, M., 2004a. Auenverbund Obere Drau—Fischo¨kologisches Monitoring – Endbericht. Studie im Auftrag des Amts der Ka¨rntner Landesregierung—Abt. 18 – Wasserwirtschaft: 94pp. Ward, J.V., Wiens, J.A., 2001. Ecotones of riverine systems: role and typology, spatio- temporal dynamics, and river regulation. Ecohydrol. Hydrobiol. 1, 25–36.

Discussion by M. Roberts Although braided rivers are a central component in the Leitbild approach to river restoration planning for gravel-bed rivers, it can be more apposite in some environments to consider using the properties of ‘wandering gravel-bed rivers’. This channel planform distinguished by multiple channels, armouring of the channel bottom and forested islands, which are often stable over decadal time periods, can be the most geomorphically appropriate planform to use in mountainous environments (Fig. 30.11). The analysis of historical maps from Alpine regions often reveals the presence of forested islands (e.g., Rhoˆne River near Aoste, France) suggesting that the wandering gravel-bed planform would be the most suitable to use for restoration work.

804

S. Muhar et al.

Figure 30.11. Geomorphic-depositional elements of a wandering gravel-bed river (Roberts et al., 1997).

Reply by the authors

In general, the Leitbild approach is not only focusing on braided rivers but takes into consideration the rivertype speciﬁc morphology (Muhar et al., 2000). The Drau river is situated in a relatively narrow valley. Based on the analysis of historic maps it could be shown, that the river is a pendulous river type with partially braiding sections, where high bedload input occurred via torrential tributaries, causing highly dynamic changes in river morphology. Even ‘‘typical braided rivers’’ in New Zealand – like the Waimakariri river on the Southern Island – are characterised by braids and islands, that are partially covered with pioneer vegetation, being stable over years (Habersack and Smart, 1999). According to the classiﬁcation schemes of Nanson and Knighton (1996) and Nanson and Croke (1992), in the pre-channalisation state, the alluvial Danube sections in Austria could be designated as gravel-dominated, laterally active anabranching river type that developed medium-energy, non-cohesive ﬂoodplains, i.e., wandering gravelbed river ﬂoodplains. Nevertheless the typical morphological character of the Danube river showed also sections with multiple channels and islands with pioneer vegetation, but within a short period of time (often less than decades) a complete turnover of bars etc. takes place (Hohensinner et al., 2004).

References Habersack, H. and Smart, G.M., 1999. Width of braided gravel bed rivers: implications for management in Austria and New Zealand, Proceedings of the IAHR-Symposium on River, Coastal and Estuarine Morphodynamics in Genova, pp. 575–584. Hohensinner, S., Habersack, H., Jungwirth, M., Zauner, G., 2004. Reconstruction of the characteristics of a natural alluvial river-ﬂoodplain system and hydromorphological changes following human modiﬁcations: the Danube River (1812–1991). J. River Res. Appl. 20, 25–41 ISSN: 1535-1467. Muhar, S., Schwarz, M., Schmutz, S., and Jungwirth, M., 2000. Identiﬁcation of rivers with high and good habitat quality: methodological approach and applications in Austria. In: Jungwirth, M., Muhar, S., and Schmutz, S. (Eds), Assessing the Ecological Integrity of Running Waters, Hydrobiologia 422/423, 343–358. Nanson, G.C., Croke, J.C., 1992. A genetic classiﬁcation of ﬂoodplains. Geomorphology 4, 459–486. Nanson, G.C., Knighton, A.D., 1996. Anabranching rivers: their cause, character and classiﬁcation. Earth Surf. Process. Landf. 21, 217–239.

Gravel bed rivers VI: from process understanding to river restoration

From Boardroom to Wedding Bed

From Boardroom to Wedding Bed

From Boardroom to Wedding Bed

The Yukon River (Rivers in World History)

The Volga River (Rivers in World History)

Understanding second language process

The Indus River (Rivers in World History)

Endovascular Aneurysm Repair: From Bench to Bed

Restoration

Restoration

Restoration

Rivers

Restoration

Restoration

Restoration

Restoration

Restoration

Restoration

Restoration

Restoration

Restoration

Restoration

Restoration

Restoration

Restoration

The Wrong Bed Early to Bed

Restoration

Restoration

Restoration

Restoration

Gravel bed rivers VI: from process understanding to river restoration

From Boardroom to Wedding Bed

From Boardroom to Wedding Bed

From Boardroom to Wedding Bed

The Yukon River (Rivers in World History)

The Volga River (Rivers in World History)

Understanding second language process

The Indus River (Rivers in World History)

Endovascular Aneurysm Repair: From Bench to Bed

Restoration

Restoration

Restoration

Rivers

Restoration

Restoration

Restoration

Restoration

Restoration

Restoration

Restoration

Restoration

Restoration

Restoration

Restoration

Restoration

Restoration

The Wrong Bed Early to Bed

Restoration

Restoration

Restoration

Restoration

Recommend Documents